Method and system for predicting bending performance of carbon fiber reinforced composite reinforced concrete filled steel tube member

Abstract

The invention provides a method and system for predicting the bending performance of a carbon fiber reinforced composite reinforced steel tube concrete member, and belongs to the field of application of high-performance composites.The method comprises the steps that a finite element model is established; wherein the finite element model is used for simulating debonding failure and fracture failure of the carbon fiber reinforced composite material in the carbon fiber reinforced composite material reinforced concrete filled steel tube member; determining key parameters based on the finite element model; generating a plurality of data sets including debonding strain and ultimate bending moment based on the key parameters; and based on the multiple data sets, a neural network model is established and trained, so that the neural network model outputs a target debonding strain and a target ultimate bending moment in the bending performance of the carbon fiber reinforced composite reinforced steel tube concrete member. Through the method provided by the invention, the debonding strain and the ultimate bending moment of the carbon fiber reinforced composite reinforced steel tube concrete member during bending can be accurately predicted.

Description

Technical Field

[0001] The present invention belongs to the field of application of high-performance composite materials, and particularly relates to a method and system for predicting the flexural performance of concrete-filled steel tubular members strengthened with carbon fiber reinforced composites. Background Art

[0002] Concrete-filled steel tubular members are widely used in building and bridge engineering due to their high load-bearing capacity and good ductility. However, with the increase of service life or the change of load conditions, such members are prone to problems of insufficient bearing capacity due to material property degradation or design defects, and thus need to be strengthened efficiently. Carbon fiber reinforced composites have become an important technical means for strengthening concrete-filled steel tubular structural members due to their high strength, light weight and convenient construction. The wet bonding method is usually adopted in the strengthening process, that is, the carbon fiber reinforced composite is bonded to the surface of the concrete-filled steel tubular member to be strengthened through an adhesive. However, during the bending process of the concrete-filled steel tubular member strengthened with carbon fiber reinforced composites, two complex failure modes may occur: debonding failure between the carbon fiber reinforced composite and the steel pipe or fracture failure of the carbon fiber reinforced composite itself. These two complex failure modes are affected by the coupling of multiple parameters such as the yield strength of steel, the number of layers of carbon fiber reinforced composites, and the bonding interface performance, resulting in extremely difficult prediction of their mechanical behavior.

[0003] The current prediction methods mainly rely on simplified empirical formulas or traditional finite element simulations. Empirical formulas usually ignore the interface nonlinear behavior and the interaction between other size and material parameters, resulting in conservative and inaccurate prediction results; although traditional finite element analysis can partially simulate the failure process, it requires a large number of trial calculations and high computational costs, and it is difficult to quickly respond to engineering needs. In addition, related solutions are mostly limited to the analysis of a single failure mode, lacking the dynamic description of the competition mechanism between debonding and fracture. Therefore, establishing an efficient and high-precision prediction method for the flexural performance of concrete-filled steel tubular members strengthened with carbon fiber reinforced composites, especially accurately quantifying the quantitative relationship between each key parameter and the debonding strain and the ultimate moment, has become the key to solving the technical bottleneck of carbon fiber reinforced composite strengthening technology. Summary of the Invention

[0004] In view of the above problems, the embodiments of the present application provide a method and system for predicting the flexural performance of concrete-filled steel tubular members strengthened with carbon fiber reinforced composites, so as to overcome the above problems or at least partially solve the above problems.

[0005] In the first aspect of the embodiments of the present application, a method for predicting the flexural performance of concrete-filled steel tubular members strengthened with carbon fiber reinforced composites is provided, including: Establishing a finite element model; wherein, the finite element model is used to simulate the debonding failure and fracture failure of the carbon fiber reinforced composite in the concrete-filled steel tubular member strengthened with the carbon fiber reinforced composite; Based on the finite element model, key parameters are determined; wherein, the key parameters include the yield strength of steel, the bonding length of carbon fiber reinforced composite material, the number of layers of carbon fiber reinforced composite material, the elastic modulus of carbon fiber reinforced composite material, the tensile strength of carbon fiber reinforced composite material, the tangential fracture energy of the bonding interface, the tangential bonding strength of the bonding interface, the diameter-thickness ratio of the steel pipe, and the diameter of the steel pipe. Based on the key parameters, multiple data sets including debonding strain and ultimate moment are generated. Based on the multiple data sets, a neural network model is established and trained so that the neural network model outputs the target debonding strain and the target ultimate moment in the flexural performance of the concrete-filled steel tube member strengthened with carbon fiber reinforced composite material.

[0006] Further, the establishment of the finite element model includes: Obtain the three-dimensional model and material parameters of the concrete-filled steel tube member strengthened with carbon fiber reinforced composite material; wherein, the material parameters at least include: steel parameters, carbon fiber reinforced composite material parameters, concrete parameters, and bonding interface parameters. Based on the three-dimensional model and the material parameters, boundary conditions and loads are applied to establish the finite element model. Wherein, the boundary conditions characterize the constraint state of the finite element model, and the loads characterize the external forces received by the finite element model.

[0007] Further, the determination of the key parameters based on the finite element model includes: Based on the finite element model, change the dimensions of the three-dimensional model and the values of the material parameters, and respectively obtain the influence degrees of the three-dimensional model and the material parameters on the performance of the concrete-filled steel tube member strengthened with carbon fiber reinforced composite material. Based on the influence degrees, determine the key parameters.

[0008] Further, the generation of multiple data sets including debonding strain and ultimate moment based on the key parameters includes: Perform parametric processing on the key parameters so that the key parameters vary within a preset range. Change the values of the key parameters, run the finite element model, and obtain the corresponding debonding strain and ultimate moment after each run of the finite element model. Associate the debonding strain and the ultimate moment obtained each time with the corresponding values of the key parameters to generate multiple data sets.

[0009] Further, establishing and training a neural network model based on the multiple datasets to enable the neural network model to output the target debonding strain and the target ultimate moment in the flexural performance of the concrete-filled steel tubular member reinforced with carbon fiber reinforced composite material includes: Performing normalization processing on the multiple datasets, and dividing the normalized multiple datasets into a training set and a test set; Constructing an initial neural network model; wherein, the initial neural network model includes: an input layer, a hidden layer, and an output layer, the input layer receives the key parameters, the output layer outputs the debonding strain or the ultimate moment, and the hidden layer is used to connect the input layer and the output layer; Based on the training set and the test set, taking minimizing the mean square error and maximizing the coefficient of determination as the solution objectives, performing multiple iterative solutions to obtain the trained neural network model; Based on the neural network model, outputting the target debonding strain and the target ultimate moment.

[0010] Further, the performing multiple iterative solutions based on the training set and the test set, taking minimizing the mean square error and maximizing the coefficient of determination as the solution objectives to obtain the trained neural network model includes the following steps: Step 1: Initializing the weights and biases in the initial neural network model; Step 2: Inputting the training set into the initial neural network model, obtaining the predicted output and performing anti-normalization processing, and determining the mean square error and the coefficient of determination based on the anti-normalized predicted output and the actual output; Step 3: According to the mean square error, updating the weights and the biases through the backpropagation algorithm to minimize the mean square error; Step 4: During the training process, using the test set to verify the neural network model to maximize the coefficient of determination; Step 5: Repeating the iterative training process of steps 2 to 4 until the neural network model reaches the preset maximum number of iterations or the model converges.

[0011] Further, the outputting the target debonding strain and the target ultimate moment based on the neural network model includes: Based on the neural network model, obtaining the updated target weights and target biases; Based on the target weights and the target biases, combining the double S-shaped transfer function and the linear transfer function, outputting the target debonding strain and the target ultimate moment.

[0012] Further, the hidden layer adopts the double S-shaped transfer function, and the output layer adopts the linear transfer function.

[0013] Further, after establishing the finite element model, the method further includes: Verifying the finite element model, wherein the verification of the finite element model includes the following steps: Comparing the calculation results of the finite element model with the test results, wherein the calculation results and the test results include the bending moment-mid-span deflection curve and the strain distribution curve; When the calculation results are consistent with the test results, it is determined that the finite element model passes the verification.

[0014] In a second aspect of the embodiments of the present application, a system for predicting the flexural performance of a concrete-filled steel tubular member reinforced with carbon fiber reinforced composite materials is provided, including: A modeling unit for establishing a finite element model; wherein the finite element model is used to simulate the debonding failure and fracture failure of the carbon fiber reinforced composite material in the concrete-filled steel tubular member reinforced with carbon fiber reinforced composite materials; A determining unit for determining key parameters based on the finite element model; wherein the key parameters include the steel yield strength, the bonding length of the carbon fiber reinforced composite material, the number of layers of the carbon fiber reinforced composite material, the elastic modulus of the carbon fiber reinforced composite material, the tensile strength of the carbon fiber reinforced composite material, the tangential fracture energy of the bonding interface, the tangential bonding strength of the bonding interface, the diameter-thickness ratio of the steel pipe, and the diameter of the steel pipe; A generating unit for generating a plurality of data sets including debonding strain and ultimate bending moment based on the key parameters; An output unit for establishing and training a neural network model based on the plurality of data sets, so that the neural network model outputs the target debonding strain and the target ultimate bending moment in the flexural performance of the concrete-filled steel tubular member reinforced with carbon fiber reinforced composite materials.

[0015] Further, the modeling unit includes: A determining unit for changing the dimensions of the three-dimensional model and the values of the material parameters based on the finite element model, and respectively obtaining the influence degrees of the three-dimensional model and the material parameters on the performance of the concrete-filled steel tubular member reinforced with carbon fiber reinforced composite materials; Determining the key parameters based on the influence degrees.

[0016] Further, the generating unit includes: A first processing unit for performing parametric processing on the key parameters so that the key parameters vary within a preset range; Change the value of the key parameter, run the finite element model, and obtain the corresponding debonding strain and ultimate bending moment after each run of the finite element model; Associate the debonding strain and the ultimate bending moment obtained each time with the value of the corresponding key parameter to generate multiple data sets.

[0017] Further, the output unit includes: A second processing unit, configured to normalize multiple data sets and divide the normalized multiple data sets into a training set and a test set; Construct an initial neural network model; wherein, the initial neural network model includes an input layer, a hidden layer, and an output layer, the input layer receives the key parameter, the output layer outputs the debonding strain or the ultimate bending moment, and the hidden layer is used to connect the input layer and the output layer; Based on the training set and the test set, with the goal of minimizing the mean square error and maximizing the coefficient of determination, perform multiple iterative solutions to obtain the trained neural network model; Based on the neural network model, output the target debonding strain and the target ultimate bending moment.

[0018] Further, the second processing unit includes the following steps: Step 1: Initialize the weights and biases in the initial neural network model; Step 2: Input the training set into the initial neural network model, obtain the predicted output and perform anti-normalization processing, and determine the mean square error and the coefficient of determination based on the predicted output and the actual output after the anti-normalization processing; Step 3: According to the mean square error, update the weights and biases through the backpropagation algorithm to minimize the mean square error; Step 4: During the training process, use the test set to verify the neural network model to maximize the coefficient of determination; Step 5: Repeat the iterative training process of steps 2 to 4 until the neural network model reaches the preset maximum number of iterations or the model converges.

[0019] Further, the second processing unit includes: A first acquisition unit, configured to obtain the updated target weights and target biases based on the neural network model; Based on the target weights and the target biases, in combination with the double S-shaped transfer function and the linear transfer function, output the target debonding strain and the target ultimate bending moment.

[0020] Further, the hidden layer adopts the double S-shaped transfer function, and the output layer adopts the linear transfer function.

[0021] Further, after establishing the finite element model, the system further includes: A verification unit, configured to verify the finite element model, wherein the verification of the finite element model includes the following steps: Comparing the calculation results of the finite element model with the test results, wherein the calculation results and the test results include a bending moment - mid-span deflection curve and a strain distribution curve; When the calculation results are consistent with the test results, it is determined that the finite element model passes the verification.

[0022] Therefore, a method for predicting the flexural performance of a concrete-filled steel tubular member reinforced with carbon fiber reinforced composite materials provided in this embodiment can simulate the complex failure modes of a concrete-filled steel tubular member reinforced with carbon fiber reinforced composite materials during the bending process by establishing a finite element model, including the debonding failure and fracture failure of the carbon fiber reinforced composite materials. Secondly, based on the finite element model, key parameters are determined. Through the yield strength of steel, the bonding length of carbon fiber reinforced composite materials, the number of layers of carbon fiber reinforced composite materials, the elastic modulus of carbon fiber reinforced composite materials, the tensile strength of carbon fiber reinforced composite materials, the tangential fracture energy of the bonding interface, the tangential bonding strength of the bonding interface, the diameter-thickness ratio of the steel pipe, and the diameter of the steel pipe in the key parameters, multiple data sets including debonding strain and ultimate bending moment can be generated, and a neural network model is established and trained using the multiple data sets. The neural network model can learn the complex non-linear relationship between the input parameters and the output results, so as to output the target debonding strain and the target ultimate bending moment of the flexural performance of a concrete-filled steel tubular member reinforced with carbon fiber reinforced composite materials. This method overcomes the deficiencies of traditional methods that require a large amount of experimental data and complex calculation processes, can accurately predict the debonding strain and ultimate bending moment of a concrete-filled steel tubular member reinforced with carbon fiber reinforced composite materials during the bending process, and solves the problem of difficult prediction of the mechanical properties of components under complex failure mode conditions. Description of the Drawings

[0023] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for the description of the embodiments of the present application will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0024] Figure 1 It is a flowchart of the steps of a method for predicting the flexural performance of a concrete-filled steel tubular member reinforced with carbon fiber reinforced composite materials provided by an embodiment of the present application; Figure 2It is a schematic diagram of a finite element model provided by an embodiment of the present application; Figure 3 It is a schematic diagram of the quantity distribution of a debonding strain finite element model provided by an embodiment of the present application; Figure 4 It is a schematic diagram of the quantity distribution of an ultimate bending moment finite element model provided by an embodiment of the present application; Figure 5 (a) It is a schematic diagram of the sensitivity analysis of debonding strain to parameter changes provided by an embodiment of the present application; Figure 5 (b) It is a schematic diagram of the sensitivity analysis of ultimate bending moment to parameter changes provided by an embodiment of the present application; Figure 6 It is a schematic diagram of the optimal number of hidden layer neurons provided by an embodiment of the present application; Figure 7 It is a schematic diagram of the structure of an initial neural network model provided by an embodiment of the present application; Figure 8 It is a schematic diagram of the training process of a neural network model provided by an embodiment of the present application; Figure 9 (a) It is for Figure 8 A schematic diagram of the performance of a debonding strain neural network model provided; Figure 9 (b) It is for Figure 8 A schematic diagram of the performance of an ultimate bending moment neural network model provided; Figure 10 (a) It is a schematic diagram of the comparison result between the debonding strain predicted by a neural network model and the target value in the test set provided by an embodiment of the present application; Figure 10 (b) It is a schematic diagram of the comparison result between the debonding strain predicted by a neural network model and the target value in the validation set provided by an embodiment of the present application; Figure 10 (c) It is a schematic diagram of the comparison result between the debonding strain predicted by a neural network model and the target value in the total set provided by an embodiment of the present application; Figure 11 (a) It is a schematic diagram of the comparison result between the ultimate bending moment predicted by a neural network model and the target value in the test set provided by an embodiment of the present application; Figure 11 (b) It is a schematic diagram of the comparison result between the ultimate bending moment predicted by a neural network model and the target value in the validation set provided by an embodiment of the present application; Figure 11 (c) It is a schematic diagram of the comparison result between the ultimate bending moment predicted by a neural network model and the target value in the total set provided by an embodiment of the present application; Figure 12Figure (a) is a schematic diagram of the comparison results of the moment - mid - span deflection curves of debonding failure of a carbon fiber reinforced composite material provided by an embodiment of the present application; Figure 12 Figure (b) is a schematic diagram of the comparison results of the moment - mid - span deflection curves of fracture failure of a carbon fiber reinforced composite material provided by an embodiment of the present application; Figure 13 Figure (a) is a schematic diagram of the comparison results of the strain distribution curves of debonding failure of a carbon fiber reinforced composite material provided by an embodiment of the present application; Figure 13 Figure (b) is a schematic diagram of the comparison results of the strain distribution curves of fracture failure of a carbon fiber reinforced composite material provided by an embodiment of the present application; Figure 14 It is a system diagram for predicting the flexural performance of a concrete - filled steel tube member reinforced with a carbon fiber reinforced composite material provided by an embodiment of the present application. Detailed implementation manners

[0025] The exemplary embodiments of the present application will be described in more detail below with reference to the accompanying drawings in the embodiments of the present application. Although the exemplary embodiments of the present application are shown in the drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided so that the present application can be more thoroughly understood and the scope of the present application can be fully conveyed to those skilled in the art.

[0026] In the related art, during the flexural process of a concrete - filled steel tube member reinforced with a carbon fiber reinforced composite material, it may experience complex failure modes, including debonding failure and fracture failure of the carbon fiber reinforced composite material. These failure modes are affected by various factors, such as material properties, geometric dimensions, bonding interface properties, etc., making it complex and challenging to predict its flexural performance. Traditional prediction methods may not fully consider all influencing factors and their interactions, resulting in inaccurate or unreliable prediction results, and may require a large amount of experimental data and complex calculation processes, with low efficiency.

[0027] In view of this, this embodiment provides a method for predicting the flexural performance of a concrete-filled steel tube member strengthened with carbon fiber reinforced composite materials. By establishing a finite element model, the mechanical behavior of the concrete-filled steel tube member strengthened with carbon fiber reinforced composite materials during the bending process is simulated, including the debonding failure and fracture failure of the carbon fiber reinforced composite materials. The finite element model can consider the influence of various factors, such as the nonlinear characteristics of materials, geometric nonlinearity, etc., so as to improve the accuracy of prediction. Then, based on the finite element model, the key parameters affecting the flexural performance of the concrete-filled steel tube member strengthened with carbon fiber reinforced composite materials are determined. These parameters include the yield strength of steel, the bonding length of carbon fiber reinforced composite materials, the number of layers of carbon fiber reinforced composite materials, the elastic modulus of carbon fiber reinforced composite materials, the tensile strength of carbon fiber reinforced composite materials, the tangential fracture energy of the bonding interface, the tangential bonding strength of the bonding interface, the diameter-thickness ratio of the steel tube, and the diameter of the steel tube. By identifying these key parameters, the flexural performance of the member can be more comprehensively understood, and based on the key parameters, multiple data sets including debonding strain and ultimate moment are generated. These data sets can be obtained through finite element analysis or experiments, providing rich data support for subsequent neural network training. Finally, using the generated data sets, a neural network model is established and trained. The neural network model can learn the complex nonlinear relationship between the input parameters (key parameters) and the output results (debonding strain and ultimate moment), so as to achieve rapid and accurate prediction of the flexural performance of the concrete-filled steel tube member strengthened with carbon fiber reinforced composite materials, and solve the above problems.

[0028] Referring to Figure 1 , Figure 1 is a flowchart of the steps of a method for predicting the flexural performance of a concrete-filled steel tube member strengthened with carbon fiber reinforced composite materials provided by an embodiment of the present application. It can be seen from Figure 1 that it includes: Step S101: Establish a finite element model; wherein, the finite element model is used to simulate the debonding failure and fracture failure of the carbon fiber reinforced composite materials in the concrete-filled steel tube member strengthened with carbon fiber reinforced composite materials.

[0029] In this embodiment, debonding failure refers to the destruction of the bonding interface between the carbon fiber reinforced composite material and the steel pipe, resulting in the loss of the bonding force between the carbon fiber reinforced composite material and the steel pipe, and the carbon fiber reinforced composite material peeling off from the surface of the steel pipe. Debonding failure usually occurs when the bonding strength between the carbon fiber reinforced composite material and the steel pipe is insufficient. Debonding failure will affect the overall bearing capacity and deformation capacity of the carbon fiber reinforced composite material strengthened concrete filled steel tubular member. Specifically, during the bending process, slippage and debonding occur at the bonding interface between the carbon fiber reinforced composite material and the steel pipe, and finally the carbon fiber reinforced composite material peels off from the surface of the steel pipe. Fracture failure refers to the fracture of the carbon fiber reinforced composite material itself, that is, the carbon fiber reinforced composite material reaches its tensile strength limit during the tensile process and fractures. Fracture failure usually occurs when the strength of the carbon fiber reinforced composite material is insufficient or there are local defects. Fracture failure will cause a sharp drop in the bearing capacity of the member, manifested as the carbon fiber reinforced composite material reaching its tensile strength limit during the bending process, showing obvious cracks and finally fracturing.

[0030] Debonding failure and fracture failure both belong to failure modes for the carbon fiber reinforced composite material strengthened concrete filled steel tubular member. In order to better predict the performance of the carbon fiber reinforced composite material strengthened concrete filled steel tubular member in practical engineering applications, it is necessary to analyze these two failure modes. Therefore, by establishing a finite element model, the occurrence process of these two failure modes can be simulated, and the influence on the performance of the carbon fiber reinforced composite material strengthened concrete filled steel tubular member can be evaluated.

[0031] In this embodiment, in order to ensure that the finite element model can detailly simulate the mechanical behavior (debonding failure and fracture failure) of the carbon fiber reinforced composite material strengthened concrete filled steel tubular member during the bending process, the established finite element model needs to consider multiple factors such as the geometric shape, material properties, bonding interface characteristics, and load conditions of the carbon fiber reinforced composite material strengthened concrete filled steel tubular member, so as to make the debonding failure and fracture failure simulated by the finite element model more accurate.

[0032] Step S102: Based on the finite element model, determine the key parameters; wherein, the key parameters include the yield strength of the steel, the bonding length of the carbon fiber reinforced composite material, the number of layers of the carbon fiber reinforced composite material, the elastic modulus of the carbon fiber reinforced composite material, the tensile strength of the carbon fiber reinforced composite material, the tangential fracture energy of the bonding interface, the tangential bonding strength of the bonding interface, the diameter-thickness ratio of the steel pipe, and the diameter of the steel pipe.

[0033] In this embodiment, through the finite element model, the specific values of each parameter are changed one by one while keeping other parameters unchanged, and the influence of each parameter change on the performance of the concrete-filled steel tube member strengthened with carbon fiber reinforced composite materials is analyzed, especially the influence on debonding failure and fracture failure. For example, increasing the yield strength of the steel can enhance the flexural capacity of the member, but may increase the debonding risk of the carbon fiber reinforced composite material. Increasing the bonding length of the carbon fiber reinforced composite material can improve the bonding force and reduce the debonding risk. Each parameter that may affect the performance of the concrete-filled steel tube member strengthened with carbon fiber reinforced composite materials is analyzed, and the parameters with significant influence on the performance are determined as key parameters, namely the yield strength of the steel, the bonding length of the carbon fiber reinforced composite material, the number of layers of the carbon fiber reinforced composite material, the elastic modulus of the carbon fiber reinforced composite material, the tensile strength of the carbon fiber reinforced composite material, the tangential fracture energy of the bonding interface, the tangential bonding strength of the bonding interface, the diameter-thickness ratio of the steel tube, and the diameter of the steel tube. In addition, based on the single-parameter analysis, the situation of multiple parameters changing simultaneously can also be considered. Specifically, by combining different parameter values, the performance of the member under the condition of multi-parameter change is simulated.

[0034] Step S103: Based on the key parameters, generate multiple data sets including debonding strain and ultimate moment.

[0035] In this embodiment, key parameters that have a significant impact on the performance of the concrete-filled steel tube member strengthened with carbon fiber reinforced composite materials are selected, namely the yield strength of the steel, the bonding length of the carbon fiber reinforced composite material, the number of layers of the carbon fiber reinforced composite material, the elastic modulus of the carbon fiber reinforced composite material, the tensile strength of the carbon fiber reinforced composite material, the tangential fracture energy of the bonding interface, the tangential bonding strength of the bonding interface, the diameter-thickness ratio of the steel tube, and the diameter of the steel tube. Then, according to the actual engineering requirements and the interpretable value range of the constitutive model, a reasonable numerical range is set for each parameter. Using the established finite element model, the specific values of each key parameter are changed one by one, and multiple finite element simulation analyses are carried out. In each simulation, key performance indicators such as debonding strain and ultimate moment during the flexural process of the member are recorded, and the debonding strain and ultimate moment data obtained from each finite element simulation are collected and sorted to form a data set. Each data set contains the debonding strain and ultimate moment values under different parameter combinations, as well as the corresponding key parameter values.

[0036] Step S104: Based on multiple data sets, establish and train a neural network model so that the neural network model outputs the target debonding strain and target ultimate moment in the flexural performance of the concrete-filled steel tube member strengthened with carbon fiber reinforced composite materials.

[0037] In this embodiment, in step S103, multiple data sets including debonding strains and ultimate bending moments under different combinations of key parameters have been obtained. Each data set includes the values of the key parameters, as well as the corresponding debonding strain and ultimate bending moment results. Then, for the multiple data sets, normalization processing is performed to convert data with different units and scales into the range of [-1, 1] to improve the training efficiency and accuracy of the neural network. Secondly, an initial neural network model is constructed. The data set is divided into a training set and a test set, and the key parameters are used as inputs, and the debonding strain and ultimate bending moment corresponding to the key parameters are used as outputs to train the initial neural network model. During the training process, the weights and bias values are iteratively optimized until the maximum number of iterations is reached or the algorithm converges. Finally, the trained neural network model is obtained, and the target debonding strain and target ultimate bending moment in the flexural performance of the concrete-filled steel tube member strengthened with carbon fiber reinforced composite materials are output using this neural network model.

[0038] In summary, by establishing a finite element model, the complex failure modes of the concrete-filled steel tube member strengthened with carbon fiber reinforced composite materials during the bending process can be simulated, including the debonding failure and fracture failure of the carbon fiber reinforced composite materials. Secondly, based on the finite element model, key parameters are determined, and by changing the yield strength of the steel, the bonding length of the carbon fiber reinforced composite material, the number of layers of the carbon fiber reinforced composite material, the elastic modulus of the carbon fiber reinforced composite material, the tensile strength of the carbon fiber reinforced composite material, the tangential fracture energy of the bonding interface, the tangential bonding strength of the bonding interface, the diameter-thickness ratio of the steel tube, and the diameter of the steel tube in the key parameters, multiple data sets including debonding strains and ultimate bending moments can be generated, and a neural network model is established and trained using the multiple data sets. The neural network model can learn the complex non-linear relationship between the input parameters and the output results, so as to output the target debonding strain and target ultimate bending moment of the flexural performance of the concrete-filled steel tube member strengthened with carbon fiber reinforced composite materials. This method overcomes the deficiencies of the traditional method that requires a large amount of experimental data and a complex calculation process, can accurately predict the debonding strain and ultimate bending moment of the concrete-filled steel tube member strengthened with carbon fiber reinforced composite materials during the bending process, and solves the problem of difficult prediction of complex failure modes.

[0039] In a specific embodiment, the establishment of the finite element model includes: obtaining the three-dimensional model and material parameters of the concrete-filled steel tube member strengthened with carbon fiber reinforced composite materials; wherein, the material parameters at least include: steel parameters, carbon fiber reinforced composite material parameters, concrete parameters, and bonding interface parameters; based on the three-dimensional model and the material parameters, boundary conditions and loads are applied to establish the finite element model; wherein, the boundary conditions characterize the constraint state of the finite element model, and the loads characterize the external forces received by the finite element model.

[0040] In this embodiment, the three-dimensional model is the size of the actual carbon fiber reinforced composite material (CFRP) strengthened concrete-filled steel tubular (CFST) member, specifically the geometric parameters such as the diameter, thickness, length of the steel pipe, the number of layers, thickness, and bonding length of the CFRP. The steel parameters may include elastic modulus, ultimate strength, yield strength, etc. The CFRP parameters may include elastic modulus, yield strength, ultimate strength, fracture energy, tensile strength, compressive strength, etc. The concrete parameters may include ultimate tensile strength, fracture energy, etc. The bonding interface parameters may include elastic modulus, fracture energy, etc. Then, according to the three-dimensional model and material parameters, a geometric model of the CFRP strengthened CFST member is established using finite element software. This geometric model includes the steel pipe, concrete, and CFRP layer. Since the boundary conditions characterize the constraint state of the finite element model, and the constraint state may include fixed constraint, displacement constraint, and symmetry constraint, the selection of boundary conditions can be set according to the actual engineering situation, which is not limited in this embodiment. The load characterizes the external force acting on the finite element model. The external force may be concentrated force, distributed force, bending moment, and pressure. When applying the load, it is determined according to the actual situation of the CFRP strengthened CFST member. For example, for a flexural member, a bending moment or lateral load can be applied to simulate its stress condition. Therefore, by applying fixed constraints to a specific part (such as one end) on the geometric model, and applying concentrated force, distributed force, and bending moment at specific positions or on specific surfaces, the boundary conditions and loads are applied to establish a finite element model.

[0041] Secondly, the CFRP strengthened CFST member may have symmetry in terms of geometric shape, material properties, and boundary conditions. Therefore, if the member is axisymmetric and sectionally symmetric, it can be selected to be cut on the symmetry plane, and 1 / 4 of the part is retained for modeling and analysis. Therefore, the construction of the finite element model can specifically refer to Figure 2 to achieve. Figure 2 is a schematic diagram of a finite element model provided by an embodiment of the present application. It can be seen from Figure 2 that considering the symmetry of geometry, material, and boundary conditions, a 1 / 4 model of the member is established. For this purpose, symmetry plane 1 and symmetry plane 2 are set. The rigid support line and loading line are used to simulate the support and loading boundaries of the specimen. The former adopts fixed constraints, and the latter releases the displacement along the y direction. The cohesive layer mesh in the end region of the CFRP is refined to simulate the debonding between the CFRP and the steel. To reduce the influence of the hourglass effect brought by this element, 4 layers of meshes are laid in the thickness direction of the steel pipe.

[0042] In a specific embodiment, determining the key parameters based on the finite element model includes: based on the finite element model, changing the dimensions of the three-dimensional model and the values of the material parameters, and respectively obtaining the influence degrees of the three-dimensional model and the material parameters on the performance of the concrete-filled steel tube member strengthened with carbon fiber reinforced composite materials; determining the key parameters based on the influence degrees.

[0043] In this embodiment, according to the finite element model, the dimensions of the three-dimensional model, such as the steel pipe diameter Ds, the steel pipe wall thickness ts, the carbon fiber reinforced composite material paste length Lcf, etc., and the material parameters, such as the steel yield strength fy, the concrete cube compressive strength fcu, the carbon fiber reinforced composite material elastic modulus Ecf, the carbon fiber reinforced composite material tensile strength fcf, etc. are defined as variable parameters. The changes in the dimensions of the three-dimensional model and the material parameters may affect the performance of the concrete-filled steel tube member strengthened with carbon fiber reinforced composite materials. Therefore, according to the actual engineering application scenario and the applicable range of material properties, the change ranges of each parameter can be set. For example: the steel pipe diameter Ds range is 100 - 1200 mm; the steel pipe diameter-thickness ratio Ds / ts range is 20 - 100; the steel pipe yield strength fy range is 200 - 800 MPa; the concrete cube compressive strength fcu range is 30 - 70 MPa; the carbon fiber reinforced composite material elastic modulus Ecf range is 100 - 500 GPa; the carbon fiber reinforced composite material tensile strength fcf range is 1000 - 5000 MPa; the number of carbon fiber reinforced composite material layers ncf range is 2 - 21; the maximum normal traction stress t1 range is 5 - 45 MPa; the pure mode I interface fracture energy GⅠ range is 0.01 - 0.2 mJ / mm2; the maximum tangential traction stress ts0 range is 10 - 30 MPa; the pure mode II interface fracture energy GⅡ range is 1 - 5 mJ / mm2; the ratio Lcf,s / Ls of the length Lcf,s of the carbon fiber reinforced composite material in the shear section to the length Ls of the member shear section ranges from 1 / 6 - 2 / 3. Based on the changes in geometric and material parameters, a large number of finite element models can be constructed, referring to Figure 3 and Figure 4 , Figure 3 is a schematic diagram of the number distribution of debonding strain finite element models provided by an embodiment of the present application, Figure 4 is a schematic diagram of the number distribution of ultimate moment finite element models provided by an embodiment of the present application. As can be seen from Figure 3 and Figure 4 , the finite element models have a wide range of geometric and material characteristics. Then, based on the debonding strain finite element model and the ultimate moment finite element model, parameter analysis is carried out to obtain the influence degrees of the three-dimensional model and the material parameters on the performance of the concrete-filled steel tube member strengthened with carbon fiber reinforced composite materials. Finally, based on the influence degrees, the parameters with high influence degrees are selected as the key parameters.

[0044] Assume that the basic parameter values are Ds = 100 mm, Ds / ts = 40, fy = 400 MPa, fcu = 40 MPa, Ecf = 200 GPa, fcf = 3000 MPa, ncf = 3, t1 = 30 MPa, GⅠ = 0.15 mJ / mm2, ts0 = 20 MPa, GⅡ = 3 mJ / mm2, and Lcf,s / Ls = 1 / 3. Refer to Figure 5 Figure (a) and Figure 5 Figure (b). Figure 5 Figure (a) is a schematic diagram of the sensitivity analysis of debonding strain to parameter changes provided by an embodiment of the present application. Figure 5 Figure (b) is a schematic diagram of the sensitivity analysis of ultimate moment to parameter changes provided by an embodiment of the present application. Figure 5 In Figure (a) and Figure 5 Figure (b), the horizontal axis represents the level of the parameter, which increases sequentially from left to right. From Figure 5 Figure (a) and Figure 5 Figure (b), it can be seen that the minimum values of t1 and GⅠ in the current parameter range are 5 MPa and 0.01 mJ / mm2 respectively, which are much lower than their normal levels of 33.4 MPa and 0.184 mJ / mm2, but still have no impact on the ultimate moment. This shows that the parameters t1 and GⅠ used to define the normal traction-separation response have little effect on the debonding strain and the ultimate moment. In addition, the debonding strain and the ultimate moment are not sensitive to the change of the concrete cube compressive strength fcu. Therefore, the maximum normal traction stress t1, the pure mode I interface fracture energy GⅠ, and the concrete cube compressive strength fcu are not used as key parameters for further analysis.

[0045] In summary, the steel yield strength, the bonding length of the carbon fiber reinforced composite material, the number of layers of the carbon fiber reinforced composite material, the elastic modulus of the carbon fiber reinforced composite material, the tensile strength of the carbon fiber reinforced composite material, the tangential fracture energy of the bonding interface, the tangential bonding strength of the bonding interface, the diameter-thickness ratio of the steel pipe, and the diameter of the steel pipe all have a greater impact on the performance of the concrete filled steel tube members strengthened by the carbon fiber reinforced composite material and can be used as key parameters.

[0046] In a specific embodiment, based on the key parameters, multiple data sets including debonding strain and ultimate moment are generated, including: performing parametric processing on the key parameters to make the key parameters change within a preset range; changing the values of the key parameters, running the finite element model, and obtaining the corresponding debonding strain and ultimate moment after each run of the finite element model; associating the debonding strain and the ultimate moment obtained each time with the values of the corresponding key parameters to generate multiple data sets.

[0047] In this embodiment, in order to ensure that the parameter ranges of the key parameters conform to actual engineering applications and avoid setting unrealistic extreme values, the key parameters are parameterized. The purpose of parameterization is to systematically vary these key parameters within a preset range so as to generate different combinations of parameter values for subsequent finite element analysis. Then, the values of the key parameters are changed within the preset range, the finite element model is run, and the corresponding debonding strain and ultimate bending moment are obtained after each run of the finite element model. Subsequently, the debonding strain and ultimate bending moment obtained each time are associated with the values of the corresponding key parameters to generate multiple data sets. One data set includes the key parameters and the debonding strain and ultimate bending moment associated with the values of the key parameters.

[0048] In a specific embodiment, based on the multiple data sets, a neural network model is established and trained to enable the neural network model to output the target debonding strain and target ultimate bending moment in the flexural performance of the concrete-filled steel tubular member strengthened with carbon fiber reinforced composite material, including: normalizing the multiple data sets and dividing the normalized multiple data sets into a training set and a test set; constructing an initial neural network model; wherein, the initial neural network model includes: an input layer, a hidden layer, and an output layer, the input layer receives the key parameters, the output layer outputs the debonding strain or the ultimate bending moment, and the hidden layer is used to connect the input layer and the output layer; based on the training set and the test set, with minimizing the mean square error and maximizing the coefficient of determination as the solution objectives, performing multiple iterative solutions to obtain the trained neural network model; based on the neural network model, outputting the target debonding strain and the target ultimate bending moment.

[0049] In this embodiment, since the different data have different units and scales, the multiple data sets are first normalized to ensure that the data in the data sets are within the same scale range, for example, restricting the variable to the range of [−1, 1]. Then, the normalized data sets are divided into a training set and a test set. For example, the training set accounts for 80% of the total data set, and the test set accounts for 20% of the total data set. The proportion of the training set and the test set allocated in this embodiment is not limited and can be set according to the actual engineering situation.

[0050] The ultimate moment and debonding strain of a concrete-filled steel tube member strengthened with carbon fiber reinforced composite materials are determined by key parameters. Due to the complex determination mechanism, a neural network model can be used to predict the ultimate moment and debonding strain of a concrete-filled steel tube member strengthened with carbon fiber reinforced composite materials. The neural network model in this embodiment can be an ANN (Artificial Neural Network) model. In the neural network model, information propagation is performed through links that receive information from processing units (neurons) and transmit it to the next neuron. Each piece of information is affected by weights, reflecting the importance of input variables to the output. Once a neuron receives information, it combines it with other information from different neurons through a combination function. Then, the combined information is transmitted to the next unit. This iterative process is repeated until the algorithm accurately fits the data or stops when the maximum number of iterations is reached. Refer to Figure 7 , Figure 7 is a schematic structural diagram of an initial neural network model provided by an embodiment of the present application. As can be seen from Figure 7 , the structure of the initial neural network model usually consists of three types of layers: an input layer, a hidden layer, and an output layer. In this embodiment, according to the training set and the test set, with minimizing the mean square error and maximizing the coefficient of determination as the solution objectives, multiple iterative solutions are performed to obtain the trained neural network model. The input layer is the input parameters for model training and testing. The hidden layer is responsible for connecting the input layer and the output layer that transmits the model results. The formula (1) for the output of the i th neuron in the hidden layer is: Formula (1); where, y i is the output of the i th hidden layer neuron, f is the activation function, is the weighted sum of the neurons, n is the number of neurons in the input layer, x j represents the input from the j th neuron, is the j rd input layer neuron to the i th hidden layer neuron weight, b i is the i th hidden layer neuron bias value.

[0051] As can be seen from Figure 7As can be seen from [reference], the input layer can input key parameters, the output layer outputs the debonding strain or the ultimate bending moment, and the hidden layer is used to connect the input layer and the output layer. Then, based on the training set and the test set, with the goal of minimizing the mean square error and maximizing the coefficient of determination, multiple iterative solutions are carried out to obtain the trained neural network model. In this embodiment, for the debonding strain and the ultimate bending moment, it cannot be guaranteed that the debonding strain and the ultimate bending moment obtained by performing multiple iterative solutions with the goal of minimizing the mean square error and maximizing the coefficient of determination are both optimal solutions. Therefore, with the goal of minimizing the mean square error and maximizing the coefficient of determination, the neural models corresponding to the debonding strain and the ultimate bending moment are trained separately to obtain the target debonding strain and the target ultimate bending moment.

[0052] Since different parameters have different units and scales, the variables (including independent variables and dependent variables) are normalized using the following formula before the training starts, and the variables are restricted to the range of [−1, 1] through formula (2).

[0053] Formula (2) Where, x is the actual value of the variable, x n is the variable x after normalization, x min and x max are the minimum and maximum values of the variable x in the database, respectively.

[0054] The neural network model is trained using the classical backpropagation (BP) method. Two common optimization algorithms, namely the Levenberg-Marquardt (L-M) and Bayesian regularization (BR) algorithms, are usually used to train the BP neural network. The L-M algorithm is a combined algorithm that combines the Newton method and the gradient descent method for smooth coordination, which can reduce the disadvantages of slow iteration speed and easy to fall into local optimum of the BP algorithm. In addition, the BR algorithm is a network training function that optimizes and updates the weight and bias values according to the L-M algorithm. This function minimizes the combination of the squared error and the weights, then determines the correct combination, and obtains a network with reasonable generalization performance.

[0055] The L-M algorithm and the BR algorithm can be used to pre-train the database separately. Since the BR algorithm does not require a validation dataset, 80% and 20% of the data can be used to train and test the neural network model, respectively. To generate the neuron output, ensure that the data is transmitted through the hidden layer and the output layer. The two use the Tansig (double S-shaped) and Purelin (linear) transfer functions respectively, as shown below: Formula (3);

[0056] Formula (4); The mean square error ( MSE ), and the coefficient of determination ( R 2 ) are used to evaluate the performance of the neural network model, and they are expressed as the following formulas respectively: Formula (5); Formula (6); Where, n ' is the total number of samples, y i is the predicted value, t i is the target value, is the mean value of the target value.

[0057] It should be noted that when Formula (5) is used to evaluate the error, the problem is that the t i with a large gap may cause the error of the neural network model to be significantly correlated with the larger t i value, which reduces the adjustment effect of the smaller t i value in the database on the parameters of the neural network model. For example, in the current database, due to considering different diameters, this results in a relatively large difference in the loads of members with large diameters ( D s = 1200 mm) and members with small diameters ( D s = 100 mm) (the maximum gap between the two is more than 100 times). This means that in the process of training the neural network model to minimize the mean square error, the training result of the model will mainly depend on the parameters of large-diameter members, while the results of small-diameter members are ignored. In other words, the current MSE method for evaluating the model error may cause the smaller t i value to be ignored when calculating the MSE value, thus significantly reducing the generalization performance of the model.

[0058] To solve the above problems, the logarithm of the load and debonding strain parameters can also be taken to reduce the influence of larger target values on the overall error. Before normalizing the debonding strain or ultimate moment in the database, Formula (7) is used to rewrite it: Formula (7); Where, ti represents the target debonding strain or the target ultimate bending moment.

[0059] Considering the issues of the number of hidden layers and neurons in the neural network model, as well as the complex non-linear situation of the model, most current neural network models only provide the prediction results of the model without providing an explicit calculation method for the prediction results. In addition, due to the use of too many neurons, the explicit calculation method provided is too complex. Therefore, under the condition of ensuring the calculation accuracy of the model, the structure of the neural network model can be defined with as few hidden layer neurons as possible. D s , t s , f y , E cf , f cf , n cf , t s0 , G Ⅱ , and L cf,s / L s parameters as inputs, and taking the ultimate bending moment prediction result , and the debonding strain prediction result as outputs. Parameters with less influence on the results, such as t 1, G Ⅰ and f cu , are ignored in the input parameters. In the prediction model of the debonding strain, the fracture failure of the carbon fiber reinforced composite material, that is, the tensile strength f cf , is ignored to ensure the debonding failure of the carbon fiber reinforced composite material. Referring to Figure 6 , Figure 6 is a schematic diagram of the optimal number of hidden layer neurons provided by an embodiment of the present application. As can be seen from Figure 6 , considering the performance of the neural network model and the complexity of the model comprehensively, the optimal numbers of hidden layer neurons in the neural network model for predicting the debonding strain and the ultimate bending moment are determined to be 6 and 5 respectively.

[0060] In a specific embodiment, based on the training set and the test set, with the goal of minimizing the mean square error and maximizing the coefficient of determination, multiple iterative solutions are performed to obtain the trained neural network model, including the following steps: Step 1: Initialize the weights and biases in the initial neural network model; Step 2: Input the training set into the initial neural network model, obtain the predicted output and perform denormalization processing, and determine the mean square error and the coefficient of determination based on the predicted output and the actual output after the denormalization processing; Step 3: According to the mean square error, update the weights and the biases through the backpropagation algorithm to minimize the mean square error; Step 4: During the training process, use the test set to verify the neural network model to maximize the coefficient of determination; Step 5: Repeat the iterative training process of Steps 2 to 4 until the neural network model reaches the preset maximum number of iterations or the model converges.

[0061] In this embodiment, the weights are the parameters connecting different neurons in the neural network, representing the importance or influence of the input signal during the transmission process, and the biases are the additional parameters of each neuron in the neural network, used to adjust the activation threshold of the neuron. Before training the initial neural network model, initialize the weights and biases in the initial neural network model. Specifically, the weights can be randomly initialized according to the number of neurons in the input layer and the hidden layer, and the biases are initialized to small random values or zero. Then, input the key parameters of the training set into the initial neural network model, calculate the predicted output of the model through forward propagation, convert the predicted output from the normalized range back to the original data range, and then compare the predicted output after denormalization with the actual output to calculate the mean square error, which can evaluate the prediction accuracy of the neural network model, and use an optimization algorithm to update the weights and biases to minimize the mean square error. During the training process, the test set can also be used to verify the model, calculate the mean square error and the coefficient of determination on the validation set to ensure the generalization ability of the neural network model and prevent overfitting. Repeat the above steps until the preset maximum number of iterations or the model convergence condition (i.e., the network output error is less than a certain threshold) is reached to complete the training and obtain the neural network model. The neural network model obtained after training can be used to predict the corresponding target debonding strain and target ultimate moment within the set parameter range.

[0062] Exemplarily, the following will refer to Figures 8 to 11 to elaborate on obtaining the target debonding strain and the target ultimate moment. Figure 8 is a schematic diagram of the training process of a neural network model provided by an embodiment of the present application. It can be seen from Figure 8 that first, let , and normalize all variables → construct the initial neural network → start the iteration count → update the weights and biases → randomly divide the dataset into a training set and a test set → train and test the constructed initial neural network model with the training set and the test set, and save the neural network model after each training until the number of iterations is greater than 1000 or the iteration converges → obtain the outputs corresponding to all neural networks and perform denormalization processing → finally select the neural network model corresponding to the minimum mean square error and the highest coefficient of determination.

[0063] Figure 9 Figure (a) is for Figure 8 a schematic diagram showing the performance of a debonding strain neural network model provided; Figure 9 Figure (b) is for Figure 8 a schematic diagram showing the performance of an ultimate bending moment neural network model provided; At the beginning of the training process, the MSE (mean square error) values of both the training set and the test set are very high. As the training progresses, the MSE values of both the training set and the test set decrease, indicating that the neural network model has learned the relationship between the input and the output. The neural network models for predicting the debonding strain and the ultimate bending moment achieved the best training results at the 130th iteration and the 389th iteration respectively, and their MSE values are 0.00035 and 0.0000158 respectively.

[0064] Figure 10 Figure (a) is a schematic diagram showing the comparison result between the debonding strain predicted by a neural network model provided in an embodiment of the present application and the target value in the test set; Figure 10 Figure (b) is a schematic diagram showing the comparison result between the debonding strain predicted by a neural network model provided in an embodiment of the present application and the target value in the validation set; Figure 10 Figure (c) is a schematic diagram showing the comparison result between the debonding strain predicted by a neural network model provided in an embodiment of the present application and the target value in the total set; Figure 11 Figure (a) is a schematic diagram showing the comparison result between the ultimate bending moment predicted by a neural network model provided in an embodiment of the present application and the target value in the test set; Figure 11 Figure (b) is a schematic diagram showing the comparison result between the ultimate bending moment predicted by a neural network model provided in an embodiment of the present application and the target value in the validation set; Figure 11 Figure (c) is a schematic diagram showing the comparison result between the ultimate bending moment predicted by a neural network model provided in an embodiment of the present application and the target value in the total set; The total set includes the test set and the validation set. From Figure 10 Figure (a), Figure 10 Figure (b), Figure 10 Figure (c), Figure 11 Figure (a), Figure 11 Figure (b) and Figure 11The comparison results in (c) show that the predicted values (Y) of all data are basically consistent with the target values (T), that is, the data points are all distributed near the line Y = T. For the neural network models predicting the debonding strain and the ultimate moment, their R 2 (coefficient of determination) values are 0.97225 and 0.99968 respectively, indicating that the trained neural network models have excellent generalization performance.

[0065] In a specific embodiment, based on the neural network model, outputting the target debonding strain and the target ultimate moment includes: based on the neural network model, obtaining the updated target weights and target biases; based on the target weights and the target biases, combining the double S-shaped transfer function and the linear transfer function, outputting the target debonding strain and the target ultimate moment.

[0066] In this embodiment, according to the neural network model, obtain the updated target weights and the deviation and the target bias .

[0067] Then, according to the target weights and the target biases, combining the double S-shaped transfer function and the linear transfer function, the explicit calculation formulas for predicting the target debonding strain and the target ultimate moment of the circular concrete-filled steel tube flexural members strengthened with carbon fiber reinforced composites can be obtained, that is, Formulas (8) and (9) are as follows: Formula (8);

[0068] Formula (9); Wherein, the calculation result of is microstrain, the unit of is MPa. In Formula (8) , i = [1, 6]; in Formula (9) , i = [1, 5]. The calculation formulas of and are as follows: Formula (10); Formula (11); Wherein, and the units of are mm; , , , the units of are MPa; G ⅡThe unit is N / mm. , , , , , , , , and , , , , , , , , , The values of... are shown in Table 1 and Table 2 respectively.

[0069] Table 1 - Constant values in the calculation equation of... ; Table 2 - Constant values in the calculation equation of... ; In a specific embodiment, after establishing the finite element model, the method further includes: Verifying the finite element model, wherein the verification of the finite element model includes the following steps: comparing the calculation results of the finite element model with the test results, wherein the calculation results and the test results include the bending moment - mid - span deflection curve and the strain distribution curve; when the calculation results are basically consistent with the test results, determining that the finite element model passes the verification.

[0070] In this embodiment, the bending moment - mid - span deflection curve and the strain distribution curve are obtained through experiments. Then, parameters consistent with the test conditions are input into the finite element model, and the finite element model is run to obtain the calculation results. Next, the calculation results of the finite element model are compared with the test results, checking the shape of the bending moment - deflection curve and key points (such as the ultimate bending moment, deflection) and the law and key points of the strain distribution (such as the maximum strain, strain gradient), and calculating the error percentage between the two. If the error is within an acceptable range (usually less than 20%), it is determined that the model passes the verification, and then the subsequent steps are carried out.

[0071] Exemplarily, refer to the comparison diagram of the calculation results of the finite element model and the test results. Figure 12 (a) is a schematic diagram of the comparison results of the bending moment - mid - span deflection curve of debonding failure of a carbon fiber - reinforced composite material provided by an embodiment of the present application; Figure 12 (b) is a schematic diagram of the comparison results of the bending moment - mid - span deflection curve of fracture failure of a carbon fiber - reinforced composite material provided by an embodiment of the present application;Figure 13 Figure (a) is a schematic diagram of the comparison results of the strain distribution curves of debonding failure of a carbon fiber reinforced composite material provided by an embodiment of the present application; Figure 13 Figure (b) is a schematic diagram of the comparison results of the strain distribution curves of fracture failure of a carbon fiber reinforced composite material provided by an embodiment of the present application; wherein, the FE result is the calculation result of the finite element model. From Figure 12 Figures (a) and Figure 12 Figures (b), it can be seen that the bending moment-mid-span deflection curve obtained by the finite element model basically coincides with the test result. From Figure 13 Figures (a) and Figure 13 Figures (b), it can be seen that the finite element model has an excellent simulation effect on the strain distribution law of the carbon fiber reinforced composite material in the concrete filled steel tube flexural member strengthened with carbon fiber reinforced composite material. The comparison results show that for the specimens with debonding failure or fracture failure of the carbon fiber reinforced composite material, the simulation results of the finite element model in terms of the bending moment-mid-span deflection curve and the strain distribution are both good. It verifies the accuracy of the finite element model in simulating the flexural performance of the concrete filled steel tube member strengthened with carbon fiber reinforced composite material.

[0072] An embodiment of the present application also provides a system for predicting the flexural performance of a concrete filled steel tube member strengthened with carbon fiber reinforced composite material. Referring to Figure 14 , Figure 14 is a system diagram of a system for predicting the flexural performance of a concrete filled steel tube member strengthened with carbon fiber reinforced composite material provided by an embodiment of the present application. From Figure 14 , it can be seen that the system includes: a building unit 141, a determining unit 142, a generating unit 143, and an output unit 144.

[0073] The building unit 141 is used to build a finite element model; wherein, the finite element model is used to simulate the debonding failure and fracture failure of the carbon fiber reinforced composite material in the concrete filled steel tube member strengthened with carbon fiber reinforced composite material; The determining unit 142 is used to determine key parameters based on the finite element model; wherein, the key parameters include the yield strength of steel, the bonding length of carbon fiber reinforced composite material, the number of layers of carbon fiber reinforced composite material, the elastic modulus of carbon fiber reinforced composite material, the tensile strength of carbon fiber reinforced composite material, the tangential fracture energy of the bonding interface, the tangential bonding strength of the bonding interface, the diameter-thickness ratio of the steel tube, and the diameter of the steel tube; The generating unit 143 is used to generate a plurality of data sets including debonding strain and ultimate bending moment based on the key parameters; The output unit 144 is used to build and train a neural network model based on the plurality of data sets, so that the neural network model outputs the target debonding strain and the target ultimate bending moment in the flexural performance of the concrete filled steel tube member strengthened with carbon fiber reinforced composite material.

[0074] In this embodiment, the establishing unit 141 includes: A determining unit, configured to change the dimensions of the three-dimensional model and the values of the material parameters based on the finite element model, and respectively obtain the influence degrees of the three-dimensional model and the material parameters on the performance of the concrete-filled steel tubular member reinforced with carbon fiber reinforced composite materials; Based on the influence degrees, determine the key parameters.

[0075] In this embodiment, the generating unit 143 includes: A first processing unit, configured to perform parametric processing on the key parameters so that the key parameters vary within a preset range; Change the values of the key parameters, run the finite element model, and obtain the corresponding debonding strain and ultimate moment each time the finite element model is run; Associate the debonding strain and the ultimate moment obtained each time with the values of the corresponding key parameters respectively, and generate a plurality of data sets.

[0076] In this embodiment, the output unit 144 includes: A second processing unit, configured to perform normalization processing on the plurality of data sets, and divide the normalized plurality of data sets into a training set and a test set; construct an initial neural network model; wherein, the initial neural network model includes: an input layer, a hidden layer, and an output layer, the input layer receives the key parameters, the output layer outputs the debonding strain or the ultimate moment, and the hidden layer is used to connect the input layer and the output layer; based on the training set and the test set, with minimizing the mean square error and maximizing the coefficient of determination as the solution objectives, perform multiple iterative solutions to obtain the trained neural network model; based on the neural network model, output the target debonding strain and the target ultimate moment.

[0077] In this embodiment, the second processing unit includes the following steps: Step 1: Initialize the weights and biases in the initial neural network model; Step 2: Input the training set into the initial neural network model, obtain the predicted output and perform anti-normalization processing, and determine the mean square error and the coefficient of determination based on the predicted output and the actual output after the anti-normalization processing; Step 3: According to the mean square error, update the weights and biases through the backpropagation algorithm to minimize the mean square error; Step 4: During the training process, use the test set to verify the neural network model to maximize the coefficient of determination; Step 5: Repeat the iterative training process of Step 2 to Step 4 until the neural network model reaches the preset maximum number of iterations or the model converges.

[0078] In this embodiment, the second processing unit includes: A first acquisition unit, configured to acquire updated target weights and target biases based on the neural network model; based on the target weights and the target biases, combine a double sigmoid transfer function and a linear transfer function to output the target debonding strain and the target ultimate bending moment.

[0079] In this embodiment, the hidden layer uses the double sigmoid transfer function, and the output layer uses the linear transfer function.

[0080] In this embodiment, after establishing the finite element model, the system further includes: A verification unit, configured to verify the finite element model, where the verification of the finite element model includes the following steps: comparing the calculation results of the finite element model with the test results, where the calculation results and the test results include a bending moment - mid - span deflection curve and a strain distribution curve; determining that the finite element model passes the verification when the calculation results are consistent with the test results.

[0081] Each embodiment in this specification is described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The same or similar parts among the embodiments can be referred to each other.

[0082] Embodiments of the present invention are described with reference to the flowcharts and / or block diagrams of methods and devices according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, and the combination of processes and / or blocks in the flowchart and / or block diagram can also be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general - purpose computer, a special - purpose computer, an embedded processor, or other programmable data - processing terminal devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data - processing terminal devices generate a device for implementing the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0083] These computer program instructions can also be stored in a computer - readable memory that can direct a computer or other programmable data - processing terminal devices to work in a specific manner, so that the instructions stored in the computer - readable memory generate a manufactured product including an instruction device, and the instruction device implements the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0084] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device, so that a series of operation steps are executed on the computer or other programmable terminal device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable terminal device provide for implementing the steps of the functions specified in one process or multiple processes and / or blocks Figure 1 One process or multiple processes and / or blocks Figure 1 Steps of the functions specified in one block or multiple blocks.

[0085] Although the preferred embodiments of the embodiments of the present invention have been described, those skilled in the art can make additional changes and modifications once they learn the basic creative concept. Therefore, the appended claims are intended to be construed as including the preferred embodiments and all changes and modifications falling within the scope of the embodiments of the present invention.

[0086] Finally, it should also be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or terminal device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or terminal device. Without further limitation, an element defined by the statement "comprising an..." does not exclude the presence of additional identical elements in the process, method, article or terminal device comprising the element.

[0087] The above has introduced in detail a method for predicting the flexural performance of a concrete-filled steel tubular member reinforced with carbon fiber reinforced composite materials provided by the present invention. Specific examples are used in this article to elaborate on the principle and implementation manner of the present invention. The description of the above embodiments is only used to help understand the method and its core idea of the present invention; at the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for predicting the flexural performance of concrete-filled steel tubular members strengthened with carbon fiber reinforced composites, characterized in that Including: Establishing a finite element model; wherein, the finite element model is used to simulate the debonding failure and fracture failure of the carbon fiber reinforced composite material in the concrete filled steel tubular member strengthened by the carbon fiber reinforced composite material; Based on the finite element model, determining key parameters; wherein, the key parameters include steel yield strength, bonding length of carbon fiber reinforced composite material, number of layers of carbon fiber reinforced composite material, elastic modulus of carbon fiber reinforced composite material, tensile strength of carbon fiber reinforced composite material, tangential fracture energy of bonding interface, tangential bonding strength of bonding interface, diameter-thickness ratio of steel pipe, and diameter of steel pipe; Based on the key parameters, generating multiple data sets including debonding strain and ultimate moment; Based on multiple said data sets, establishing and training a neural network model so that the neural network model outputs the target debonding strain and target ultimate moment in the flexural performance of the concrete filled steel tubular member strengthened by the carbon fiber reinforced composite material.

2. The method according to claim 1, wherein The establishing of the finite element model includes: Obtaining the three-dimensional model and material parameters of the concrete filled steel tubular member strengthened by the carbon fiber reinforced composite material; wherein, the material parameters at least include: steel parameters, carbon fiber reinforced composite material parameters, concrete parameters, and bonding interface parameters; Based on the three-dimensional model and the material parameters, applying boundary conditions and loads to establish the finite element model; wherein, the boundary conditions characterize the constraint state of the finite element model, and the loads characterize the external forces received by the finite element model.

3. The method according to claim 2, wherein The determining of the key parameters based on the finite element model includes: Based on the finite element model, changing the dimensions of the three-dimensional model and the values of the material parameters, and respectively obtaining the influence degrees of the three-dimensional model and the material parameters on the performance of the concrete filled steel tubular member strengthened by the carbon fiber reinforced composite material; Based on the influence degrees, determining the key parameters.

4. The method according to claim 1, wherein The generating of multiple data sets including debonding strain and ultimate moment based on the key parameters includes: Performing parametric processing on the key parameters to make the key parameters change within a preset range; Changing the values of the key parameters, running the finite element model, and obtaining the corresponding debonding strain and ultimate moment after each running of the finite element model; Associating the debonding strain and ultimate moment obtained each time with the corresponding values of the key parameters to generate multiple said data sets.

5. The method according to claim 1, wherein The establishing and training of the neural network model based on multiple said data sets so that the neural network model outputs the target debonding strain and target ultimate moment in the flexural performance of the concrete filled steel tubular member strengthened by the carbon fiber reinforced composite material includes: Normalizing multiple said data sets and dividing the normalized multiple data sets into a training set and a test set; Constructing an initial neural network model; wherein, the initial neural network model includes: an input layer, a hidden layer, and an output layer, the input layer receives the key parameters, the output layer outputs the debonding strain or the ultimate moment, and the hidden layer is used to connect the input layer and the output layer; Based on the training set and the test set, taking minimizing the mean square error and maximizing the coefficient of determination as the solution objectives, perform multiple iterative solutions to obtain the trained neural network model; Based on the neural network model, output the target debonding strain and the target ultimate moment.

6. The method according to claim 5, wherein The process of taking minimizing the mean square error and maximizing the coefficient of determination as the solution objectives based on the training set and the test set, performing multiple iterative solutions to obtain the trained neural network model includes the following steps: Step 1: Initialize the weights and biases in the initial neural network model; Step 2: Input the training set into the initial neural network model, obtain the predicted output and perform anti-normalization processing, and determine the mean square error and the coefficient of determination based on the predicted output and the actual output after the anti-normalization processing; Step 3: According to the mean square error, update the weights and the biases through the backpropagation algorithm to minimize the mean square error; Step 4: During the training process, use the test set to verify the neural network model to maximize the coefficient of determination; Step 5: Repeat the iterative training process of Step 2 to Step 4 until the neural network model reaches the preset maximum number of iterations or the model converges.

7. The method according to claim 5, wherein The process of outputting the target debonding strain and the target ultimate moment based on the neural network model includes: Based on the neural network model, obtain the updated target weights and target biases; Based on the target weights and the target biases, combine the double S-shaped transfer function and the linear transfer function to output the target debonding strain and the target ultimate moment.

8. The method according to claim 7, wherein The hidden layer uses the double S-shaped transfer function, and the output layer uses the linear transfer function.

9. The method according to claim 1, wherein After establishing the finite element model, the method further includes: Verify the finite element model, where the verification of the finite element model includes the following steps: Compare the calculation results of the finite element model with the test results, where the calculation results and the test results include the moment-midspan deflection curve and the strain distribution curve; When the calculation results are consistent with the test results, determine that the finite element model passes the verification.

10. A system for predicting the flexural performance of concrete-filled steel tubular members strengthened with carbon fiber reinforced composites, characterized in that, Include: A modeling unit for establishing a finite element model; where the finite element model is used to simulate the debonding failure and fracture failure of the carbon fiber reinforced composite material in the carbon fiber reinforced composite material strengthened concrete filled steel tubular member; A determination unit for determining key parameters based on the finite element model; where the key parameters include the steel yield strength, the bonding length of the carbon fiber reinforced composite material, the number of layers of the carbon fiber reinforced composite material, the elastic modulus of the carbon fiber reinforced composite material, the tensile strength of the carbon fiber reinforced composite material, the tangential fracture energy of the bonding interface, the tangential bonding strength of the bonding interface, the steel pipe diameter-thickness ratio, and the steel pipe diameter; A generation unit for generating multiple data sets including debonding strain and ultimate moment based on the key parameters; An output unit for establishing and training a neural network model based on the multiple data sets, so that the neural network model outputs the target debonding strain and the target ultimate bending moment in the flexural performance of the concrete-filled steel tubular member strengthened with carbon fiber reinforced composite materials.

Images