Reliability analysis method for axial impact of high-strength steel pipe ultra-high performance concrete member

Abstract

The invention relates to the technical field of building and civil engineering, and discloses a high-strength steel pipe ultra-high performance concrete member axial impact reliability analysis method, which comprises: collecting a high-strength steel pipe ultra-high performance concrete member axial impact sample set, each axial impact sample comprising an impact load parameter, a geometric parameter and a material parameter of the member; based on the axial impact sample set, an Ann neural network is trained, and a residual bearing capacity coefficient prediction model is constructed; on the basis of the predicted residual bearing capacity coefficient, the damage grade of the component is obtained, and a limit performance function of the damage grade of the component is established by setting a residual bearing capacity coefficient threshold and combining geometric parameters and material parameters of the component so as to predict the damage reliability and the damage probability of the component; according to the structural safety level, a reliability threshold value is determined, and whether the component meets the performance requirement under impact or not is judged, so that component performance design is achieved; the method improves the prediction precision of the residual bearing capacity coefficient of the component, and can effectively carry out reliability analysis on the damage of the component.

Description

Technical Field

[0001] The present invention relates to the technical field of construction and civil engineering, and in particular to a reliability analysis method for axial impact of high-strength steel tube ultra-high performance concrete components. Background Art

[0002] Ultra-high-performance concrete filled high-strength steel tube (UHPC-FHST) components align with the development trend of modern architecture towards green, lightweight, high-performance, and economical construction. The combination of these two high-performance materials gives UHPC-FHST exceptional mechanical, seismic, and durability properties, making this composite structure widely used in engineering structures such as large industrial plants, high-rise buildings, bridges, and long-span transmission towers. In addition to conventional static loads, impact loads pose a significant safety risk to UHPC-FHST components, characterized by diverse sources, low duration, and high hazard potential. Impact loads can be categorized into lateral and axial impacts based on load properties. Lateral impacts primarily originate from the impact of runaway vehicles, derailed trains, and ships on the structure. Axial impact scenarios include the impact of falling objects such as collapsed rocks, billboards, and decorative materials on bridge piers, helicopter crashes, and building collapses caused by severe fires or explosions. Severe axial impacts severely destroy the bearing capacity of structural columns, which may further lead to more serious overall structural collapse. Therefore, proper design and strengthening of the structure's resistance to axial impact remain challenges facing structural engineers.

[0003] Axial impact and post-impact static loading scenarios such as Figure 9 As shown in the figure, a mass block with mass M and initial velocity V hits a square UHPC-FHST column with side length B and column length L. The thickness and yield strength of the steel pipe are t and f respectively. y , the compressive strength of concrete cube is f cu For high-strength UHPC-FHST components, impact loads may not completely destroy their bearing capacity. They can retain some residual axial bearing capacity, and the residual bearing capacity coefficient β = P r / P u , where P u and P r They are the initial axial bearing capacity and the residual bearing capacity after impact of UHPC-FHST, respectively.

[0004] As a load-bearing component, UHPC-FHST with slight or moderate damage has the potential to continue bearing loads after reinforcement. Therefore, quantifying the residual performance of components after axial impact is crucial for quickly assessing damage and providing removal or repair recommendations. However, due to factors such as geometric parameters, material properties, and nonlinear interactions, the calculation of the bearing capacity of steel tube concrete is relatively complex. In particular, the impact is accompanied by concrete cracks and steel tube buckling, which indicates that the use of classical mechanical models to establish the objective function of the residual bearing capacity of UHPC-FHST still faces challenges. At the same time, in practice, the geometric dimensions, material properties, and loads of the components are all discrete to a certain extent. In this case, the response of the structure will show a certain degree of randomness. Therefore, structural reliability analysis based on probabilistic analysis becomes key, which has not been reported in current research.

[0005] Existing experiments or finite element analysis have focused on parametric research and mechanistic analysis of the dynamic behavior of UHPC-FHST columns subjected to axial impact. The research results can be summarized as follows: 1) The dynamic response of UHPC-FHST columns under axial impact is affected by the impact energy, cross-sectional dimensions, steel tube thickness, and material properties; 2) The failure mode of UHPC-FHST columns under axial impact is similar to that under static load, both manifesting as local buckling of the steel tube; 3) Improving steel strength and cross-sectional steel content can significantly enhance the residual bearing capacity of UHPC-FHST columns after impact while reducing axial deformation; 4) Existing research provides a method for calculating the residual bearing capacity coefficient β of UHPC-FHST columns after axial impact, namely:

[0006]

[0007] Among them, u m is the maximum axial deformation; L is the column length; E is the impact energy; m * =m c / M,m c and M are the mass of UHPC-FHST and the mass of impact body respectively; v * =V / 10, V is the impact velocity; A is the cross-sectional area of ​​UHPC-FHST; ξ is the steel pipe constraint factor; f cu When β is less than 100MPa c Take 1.0, otherwise take 0.9; f c is the compressive strength of the concrete prism; f y Indicates the yield strength of steel pipe.

[0008] The disadvantages of the above method are:

[0009] (1) Insufficient calculation accuracy and effectiveness

[0010] Existing research uses formula When establishing the calculation method of the residual bearing capacity coefficient β, the β of 420 random samples and the maximum axial displacement um The scatter plot distribution of β and u m It is roughly expressed as an exponential function distribution relationship, so based on these 420 samples, the formula is proposed and formula The u shown m and β regression formula. Although this method is effective for axial displacement u m A higher precision fitting is achieved, but the fitting of β is done at u m It is not directly established by parameters such as impact load, UHPC-FHST geometry and material strength, which leads to the formula The calculation accuracy of β is insufficient, making it difficult to accurately reflect the residual capacity of UHPC-FHST after axial impact.

[0011] (2) Failure to conduct reliability analysis on damage of UHPC-FHST

[0012] Influenced by factors such as manufacturing process, construction errors, and external environment, the cross-sectional dimensions and material properties of UHPC-FHST columns in actual scenarios can be regarded as random variables. Structural design in such uncertain scenarios should be based on probabilistic analysis, and there is a lack of damage reliability analysis of UHPC-FHST based on residual bearing capacity. Summary of the Invention

[0013] In response to the above-mentioned deficiencies in the prior art, the present invention provides a reliability analysis method for axial impact of high-strength steel tube ultra-high performance concrete components, which is used to solve the defects of inaccurate calculation of the residual bearing capacity of components and insufficient effectiveness in the prior art, and at the same time realize reliability analysis of component damage.

[0014] In order to achieve the above-mentioned object of the invention, the technical solution adopted by the present invention is:

[0015] The reliability analysis method of high-strength steel tube ultra-high performance concrete components under axial impact includes the following steps:

[0016] S1. Collect axial impact sample sets of high-strength steel tube ultra-high performance concrete components. Each axial impact sample includes the impact load parameters, geometric parameters, and material parameters of the component.

[0017] S2. Based on the axial impact sample set, the Ann neural network is trained and a residual bearing capacity coefficient prediction model is constructed to obtain the predicted residual bearing capacity coefficient;

[0018] S3. Based on the predicted residual bearing capacity coefficient, the damage level of the component is obtained. By setting a threshold of the residual bearing capacity coefficient and combining the geometric parameters and material parameters of the component, a limit function of the component damage level is established to predict the component damage reliability and damage probability;

[0019] S4. Determine the reliability threshold based on the structural safety level to determine whether the component meets the performance requirements under impact, so as to achieve component performance design.

[0020] The present invention has the following beneficial effects:

[0021] 1. The reliability analysis method for high-strength steel tubular ultra-high performance concrete components under axial impact proposed in this paper introduces the sparrow search method into the Ann neural network to establish a hybrid machine learning model, namely the SSA-Ann neural network model, to learn and train the residual bearing capacity coefficient of UHPC-FHST components under axial impact, thereby improving the model stability and the prediction accuracy of the component residual bearing capacity coefficient;

[0022] 2. Based on the training model, a prediction model for the residual bearing capacity coefficient of UHPC-FHST components after impact was established, which greatly facilitates the application of designers and improves work efficiency;

[0023] 3. The established database covers almost all the material strength, structural geometry, and impact load parameter ranges of high-performance steel tube ultra-high performance concrete columns, making the established residual bearing capacity coefficient prediction model formula widely applicable;

[0024] 4. Based on the application characteristics of UHPC-FHST components, the predicted residual bearing capacity coefficient is used to evaluate the damage level of UHPC-FHST components after impact. A calculation method for component damage reliability and damage probability is further established. The reliability is based on the random distribution of input parameter variables to ensure that the reliability threshold can be determined for each damage level according to the structural safety level during actual design. Finally, the axial impact resistance design of UHPC-FHST components is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 Schematic diagram of the process of the reliability analysis method for axial impact of high-strength steel tube ultra-high performance concrete components proposed in the present invention;

[0026] Figure 2 Schematic diagram of the fitness curve of the iterative training of the Ann neural network in the embodiment;

[0027] Figure 3 This is a schematic diagram showing the scatter comparison of the results of using SSA-Ann and Ann to predict UHPC-FHST components under the training set in the embodiment;

[0028] Figure 4 This is a schematic diagram showing the scatter comparison of the results of using SSA-Ann and Ann to predict UHPC-FHST components under the test set in the embodiment;

[0029] Figure 5 Schematic diagram of the structure of the SSA-Ann neural network model in the embodiment;

[0030] Figure 6 Schematic diagram of the comparison results of the residual bearing capacity coefficient calculated using the method proposed by the present invention and the existing method under the training set in the embodiment;

[0031] Figure 7 Schematic diagram showing the comparison results of the residual bearing capacity coefficient calculated using the method proposed by the present invention and the existing method under the test set in the embodiment;

[0032] Figure 8 This is a schematic diagram showing the ranking of correlation coefficients between input parameters and damage probability in the embodiment;

[0033] Figure 9 Schematic diagram of axial impact and static loading of UHPC-FHST components. DETAILED DESCRIPTION

[0034] The specific embodiments of the present invention are described below to facilitate understanding of the present invention by those skilled in the art. However, it should be clear that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the appended claims, these changes are obvious, and all inventions and creations utilizing the concepts of the present invention are protected.

[0035] like Figure 1 As shown, the reliability analysis method for axial impact of high-strength steel tube ultra-high performance concrete components includes the following steps S1-S4:

[0036] S1. Collect axial impact sample sets of high-strength steel tube ultra-high performance concrete components. Each axial impact sample includes the impact load parameters, geometric parameters, and material parameters of the component.

[0037] Specifically, the impact load parameters of the component include impact mass and impact velocity; the geometric parameters include column length, square section side length, and steel pipe thickness; and the material parameters include steel pipe yield strength and cubic concrete compressive strength.

[0038] In this embodiment, in order to effectively train the artificial intelligence model, it is necessary to build a comprehensive database. Existing studies have used ABAQUS / Explicit software to numerically simulate the impact performance and residual bearing capacity of UHPC-FHST columns under axial impact loads, and provided 480 open source samples. Therefore, the present invention collects these samples for machine learning training; among them, the present invention uses the numerical model to compare and calibrate the failure mode and axial compression deformation-load curve of 50 experimental samples, and finds that the validity and reliability of these samples are trustworthy. Therefore, by collecting these samples, a high-strength steel tube ultra-high performance concrete component axial impact sample set is generated, in which each sample input variable includes 7 parameters, which can be roughly divided into impact load (impact mass M, impact velocity V), UHPC-FHST geometric parameters (column length L, square section side length B, steel tube thickness t) and material parameters (steel tube yield strength f y and cubic concrete compressive strength f cu ); Existing research, through experiments and finite element analysis, has shown that the above seven factors are key factors affecting the axial impact response of UHPC-FHST components. Therefore, this invention uses these seven factors as input parameter variables. Table 1 provides the statistical information of the input parameters, as shown below:

[0039] Table 1 Input parameter statistics

[0040]

[0041]

[0042] Table 1 above summarizes the statistical information of the input parameters of the components, including minimum, maximum, mean, standard deviation, and coefficient of variation. The input parameters of these samples are constructed based on actual working conditions and impact scenarios, covering a wide range of parameters, including impact parameters (M = 0-16t, V = 0-20m / s, impact energy E = 0-3000kJ, impact momentum I = 0-300t·m / s), geometric dimensions (L = 1700-4300mm, B = 400-900mm, t = 7-26mm, aspect ratio L / B = 2.83-5, width-to-thickness ratio B / t = ​​23.08-100), material properties (f y =235-960MPa, f cu =104-199 MPa). It should be noted that this invention does not involve the elastic modulus of concrete. This is because the elastic modulus of concrete is generally determined by its compressive strength. Since all specimens have compressive strength values, the redundancy of the elastic modulus input feature can be reduced. Similarly, the input parameters do not involve the ultimate stress of steel.

[0043] S2. Based on the axial impact sample set, the Ann neural network is trained and a residual bearing capacity coefficient prediction model is constructed to obtain the predicted residual bearing capacity coefficient.

[0044] Specifically, based on the axial impact sample set, the specific process of training the Ann neural network is as follows:

[0045] Normalize the parameters of each axial impact sample in the axial impact sample set to obtain a normalized axial impact sample, namely:

[0046]

[0047] Among them, x ip represents the normalized value of the i-th parameter of the axial impact sample, x i Represents the original value of the i-th parameter of the axial impact sample, x i,min Indicates the minimum value of the i-th parameter of the axial impact sample, x i,max Indicates the maximum value of the i-th parameter of the axial impact sample.

[0048] In this embodiment, since data quality affects the model training performance, all data are randomly arranged and then the data set is divided into an 8:2 ratio, that is, 384 and 96 samples are used as the training set and test set, respectively; for the training set and test set, the input parameters are the above 7 parameters, and the output parameter is the residual bearing capacity coefficient.

[0049] Based on the normalized axial impact samples, the Ann neural network is trained and a residual bearing capacity coefficient prediction model is constructed.

[0050] Specifically, based on the normalized axial impact samples, the specific process of training the Ann neural network is as follows:

[0051] Initialize the hyperparameters of the Ann neural network, including the number of hidden layers, the number of hidden layer neurons, different types of hidden layer transfer functions, and the weights and biases corresponding to each transfer function.

[0052] Different types of hidden layer transfer functions and their corresponding weights and biases are used as inputs of the sparrow search method. By setting the population number, number of iterations, upper and lower limits of weights and biases, an iterative search is performed to obtain the optimal weights and biases.

[0053] The transfer function corresponding to the optimal weight and bias is used as the optimal transfer function of the hidden layer.

[0054] Specifically, the optimal transfer function of the hidden layer is a logarithmic transfer function.

[0055] Based on the optimal transfer function of the hidden layer and its corresponding optimal weights and biases, the normalized axial impact samples are input into the Ann neural network for several training times, and finally a trained SSA-Ann neural network model is generated.

[0056] In this embodiment, the initial network weights and biases, the number of hidden layers, the number of hidden layer neurons, and the type of hidden layer transfer function of the Ann neural network model are all model hyperparameters, which directly affect the prediction accuracy, generalization performance, and training efficiency of the model. The Sparrow Search Algorithm (SSA) is a heuristic optimization method based on the foraging and anti-predation behavior of sparrows. The SSA population can be divided into discoverers, joiners, and scouts according to the proportion of individuals. The sparrow individuals who find better food are discoverers, and the scouts are responsible for scouting the enemy. The remaining sparrows are joiners. After the scouts find danger, the sparrow population abandons the food. Therefore, the present invention uses SSA to determine the optimal initial network weights and biases of the Ann neural network. The hybrid model integrated with SSA and Ann (which can be simplified to SSA-Ann neural network model) is used to predict the residual bearing capacity coefficient of UHPC-FHST components under axial impact. The specific process is: 1) Determine the number of hidden layer neurons L1; the number of hidden layers is taken as 1, so as to derive a function expression that is easy to calculate, which can be calculated according to the formula L1=log2a, Initially set the L1 range, where a is the input parameter (taken as 7), b ′ is the number of output parameters (taken as 1), α ′ is a constant ranging from 1 to 10. Therefore, the L1 range is set to 4 to 12. Through multiple trials, it was found that the optimal value of L1 is 9, which can simultaneously capture nonlinear relationships and avoid overfitting risks; 2) SSA finds the optimal initial weights and biases; the upper and lower limits of weights and biases are set to 1.5 and -1.5 respectively, the number of sparrow populations is 30, and the SSA method is iterated 1000 times to find the initial optimal weights and biases of the hidden layer transfer function, that is, the optimal initial weights and biases of the linear transfer function Purelin, the logarithmic transfer function Logsig and the tangent transfer function Tansig, and the root mean square error RMSE is used as the fitness, and the weights and biases corresponding to the minimum fitness are identified as the optimal network configuration under the current iteration; among them, the iterative curve is as follows Figure 2As shown in the figure, when the number of iterations is 1000, the fitness is the smallest when the logarithmic transfer function Logsig is used, so the hidden layer uses the Logsig transfer function, and the corresponding network weights and biases are the optimal initial network weights and biases; 3) Substitute into the Ann neural network training; the optimal initial network weights and biases are brought into the neural network training for 500 times, and the hidden layer still uses the Logsig transfer function, and the learning rate is as small as possible, which is 0.01. Finally, the prediction results and the final trained SSA-Ann neural network weights and biases are output, and the model performance is evaluated using five indicators, namely: correlation coefficient (R 2 ), root mean square error (RMSE), mean absolute error (MAE), average value of the ratio of predicted value to actual value (MEAN), coefficient of variation (MEAN), and their corresponding formulas are: n represents the number of data points, P j 、T j They represent the predicted value of the residual bearing capacity coefficient of the jth data and the true value of the residual bearing capacity coefficient, P avg 、T avg Represent the average predicted value and the average true value respectively; that is, these indicators can comprehensively reflect the reliability of the model in terms of prediction relevance, absolute error, and prediction robustness; the optimal value of R is 1.0 2 The RMSE, MAE, and COV values ​​of 0 indicate that the forecast results are in perfect agreement with the true values, while the RMSE, MAE, and COV values ​​of 0 indicate the smallest forecast error and forecast discreteness.

[0057] like Figure 3-Figure 4 As shown, Figure 3 、 Figure 4 The scatter comparison results of the residual bearing capacity coefficient of UHPC-FHST components predicted by SSA-Ann and Ann under the training set and test set are shown respectively, among which β pre and β true are the predicted values ​​and the true values ​​respectively; three oblique lines are provided in each figure: y = 1.2x, y = 0.8x and 45° oblique line. The results show that the prediction accuracy of SSA-Ann exceeds that of Ann; that is, for the training set, the correlation coefficient R between SSA-Ann and Ann is 2 The R values ​​of SSA-Ann and Ann are 0.90 and 0.83, respectively, and the root mean square error (RMSE) values ​​are 0.09 and 0.07, respectively. 2The scatter points in the SSA-Ann graph are almost evenly distributed relative to the 45° line, showing an unbiased prediction of the residual bearing capacity. The data points in the training set and test set are distributed within the range of y = 1.2x and y = 0.8x, accounting for 91.41% and 88.54% respectively. This shows that the developed hybrid machine learning model has achieved an accurate prediction of the residual behavior of UHPC-FHST components after impact. The SSA-Ann neural network model of the final component, that is, the Ann neural network structure with the introduction of the sparrow search method, is shown in Figure 2. Figure 5 As shown, it includes an input layer, a hidden layer and an output layer. The transfer function of the hidden layer is a logarithmic transfer function Logsig, and in order to derive a simpler calculation formula, the transfer function of the output layer selects a linear transfer function Purelin.

[0058] Specifically, the process of constructing the residual bearing capacity coefficient prediction model is as follows:

[0059] The normalized axial impact sample is input into the input layer of the trained SSA-Ann neural network model, and the first signal matrix is ​​output to the hidden layer. The first signal matrix received by the hidden layer is:

[0060] α=xν+γ

[0061] Where α represents the first signal matrix, x represents the normalized axial impact sample matrix, ν and γ represent the weight matrix and bias matrix between the input layer and the hidden layer, respectively.

[0062] In this embodiment, Figure 5 As shown in Figure 2, the seven input characteristics of the normalized axial impact sample, namely, impact load (impact mass M, impact velocity V), UHPC-FHST geometric parameters (column length L, square section side length B, steel tube thickness t) and material parameters (steel tube yield strength f y and cubic concrete compressive strength f cu ) are respectively input into the input layer of the SSA-Ann neural network model, and the final input layer outputs the first signal matrix.

[0063] According to the optimal transfer function of the hidden layer, the second signal matrix is ​​output to the output layer, and the second signal matrix received by the output layer is:

[0064] b=Logsig(α)

[0065] Wherein, b represents the second signal matrix, and Logsig represents the logarithmic transfer function.

[0066] The output signal of the final output layer is:

[0067] λ=Purelin(bw+θ)=bw+θ

[0068] Among them, λ represents the output signal, w and θ represent the weight matrix and bias matrix between the hidden layer and the output layer, respectively.

[0069] According to the output signal, the residual bearing capacity coefficient prediction model is constructed, namely:

[0070]

[0071] Among them, β max , β min Respectively represent the maximum and minimum values ​​of the true value of the residual bearing capacity coefficient, β pre Represents the predicted residual bearing capacity coefficient.

[0072] In this embodiment, since the output signal λ is a normalized parameter, the predicted residual bearing capacity coefficient should be denormalized to construct the above-mentioned residual bearing capacity coefficient prediction model. In addition, the network parameters ν, γ, w, and θ (retain two decimal places) saved after the SSA-Ann neural network model training is completed are as follows:

[0073]

[0074] γ=[-21.38 1.74 -25.11 12.35 -1.96 1.17 -32.17 5.38 3.79]

[0075] w=[14.12 -2.48 -12.03 -0.47 -2.32 0.42 16.53 -0.27 2.85]T

[0076] θ=[0.97]

[0077] Where T represents the matrix transpose.

[0078] Therefore, the specific values ​​of the above network parameters ν, γ, w, and θ are substituted into the residual bearing capacity coefficient prediction model, and each term in the equation is retained to 3 significant digits. The expression of the residual bearing capacity coefficient prediction model can be obtained as follows:

[0079]

[0080] Where e is an exponential function, and y1 to y9 are linear combinations of the seven input parameters in the hidden layer, which are expressed in matrix form as follows (retaining three significant digits):

[0081]

[0082] In addition, the expression of the residual bearing capacity coefficient prediction model for β preThe regression does not involve any physical background, and β pre When it exceeds 1.0, β pre Take 1.0. It can be inferred that once the β of the UHPC-FHST component after impact is obtained through the developed expression pre , combined with the initial bearing capacity P provided by the Technical Specification for Concrete-Filled Steel Tube Structures (GB50936-2014) u , designers can quickly evaluate the residual capacity P of components r .in, Figure 6-Figure 7 The formula proposed by the present invention and the existing method (i.e., formula ) to calculate the residual bearing capacity coefficient, since the existing method does not consider the scenario where the impact velocity and impact mass are 0, Figure 6-Figure 7 Only the samples with impact mass and velocity greater than 0 are plotted for comparison, where the training set and test set have 336 and 85 data points respectively; Figure 6-Figure 7 It can be seen that the calculation accuracy of the proposed method is significantly higher than that of the existing method, and the MEAN of the training set and test set calculated by the present invention are 1.01 and 1.02, respectively, which are close to 1.0; 89.88% and 87.06% of the data in the training set and test set are distributed in the regions y = 1.2x and y = 0.8x, respectively; while when calculated using the existing method, 36.01% and 43.53% of the data in the training set and test set are distributed in the regions y = 1.2x and y = 0.8x, respectively.

[0083] S3. Based on the predicted residual bearing capacity coefficient, the damage level of the component is obtained. By setting the residual bearing capacity coefficient threshold and combining the geometric parameters and material parameters of the component, the limit function function of the component damage level is established to predict the component damage reliability and damage probability.

[0084] Specifically, step S3 includes S31-S35:

[0085] S31. Based on the predicted residual bearing capacity coefficient, the components are classified into damage levels, including mild damage, moderate damage and severe damage.

[0086] S32. Obtain the coefficient of variation of the geometric parameters and material parameters of the component, where the geometric parameters and material parameters of the component are normally distributed random variables.

[0087] In this embodiment, the geometric parameters (side length of the square section, thickness of the steel pipe) and material parameters (yield strength of the steel pipe, compressive strength of the cubic concrete) have a coefficient of variation and can be regarded as random variables to obtain the limiting function function in the next step.

[0088] S33. Input the geometric parameters and material parameters of the component, and establish the limit function of the component damage level by setting the residual bearing capacity coefficient threshold, that is:

[0089] G(B,t,f y ,f cu )=β pre -[β]

[0090] Among them, G represents the limit function of component damage level, B represents the side length of the square section, t represents the thickness of the steel pipe, and f t Indicates the yield strength of the steel pipe, f cu represents the compressive strength of cubic concrete, and [β] represents the threshold value of the residual bearing capacity coefficient.

[0091] S34. Based on the limit function of component damage level, the first-order second moment method is used to calculate the component damage reliability, that is:

[0092]

[0093] Among them, ψ represents the component damage reliability, σ B , σ t 、 They represent the standard deviations of the side length of the square section, the thickness of the steel pipe, the yield strength of the steel pipe, and the compressive strength of the cubic concrete, respectively, and the standard deviations are obtained by multiplying the corresponding mean values ​​by the coefficient of variation.

[0094] S35. Calculate the component damage probability based on the component damage reliability, namely:

[0095] P=φ(-ψ)

[0096] Where P represents the component damage probability, and φ represents the cumulative function of the normal distribution function.

[0097] In this embodiment, UHPC-FHST is used as a vertical load-bearing component, based on the predicted residual bearing capacity coefficient β pre The impact damage level of UHPC-FHST components is divided into mild damage (0.8<β pre <0.95), moderate damage (0.5<β pre <0.8) and severe injury (β pre <0.5). In practice, the geometric dimensions, material properties, and loads of components are all discrete to a certain extent. At this time, the structural response will show a certain randomness. Therefore, the structural reliability analysis based on the probability model becomes the key. Referring to existing studies, the statistical distribution of component dimensions and material properties is assumed to be a normal distribution. Therefore, the square section side length B, steel pipe thickness t, steel pipe yield strength f of the UHPC-FHST component proposed in this invention are y and concrete compressive strength f cuThe COV of the standard normal distribution series are 0.02, 0.02, 0.05, and 0.11 respectively.

[0098] At the same time, the factors affecting the structural reliability include the load effect S and resistance R of the component, and in this invention, R can be recorded as β pre , S is the residual bearing capacity coefficient threshold [β] (and for mild, moderate and severe damage, [β] is 0.95, 0.8 and 0.5 respectively), so the limit function of the damage level of UHPC-FHST components is expressed as G(B,t,f y ,f cu )=β pre -[β]; where B, t, and f y 、f cu are the four random parameter variables of the function, β pre It can be obtained by substituting the above expression of the residual bearing capacity coefficient prediction model into specific values; finally, the first-order second moment method is used to solve the limit function to obtain the component reliability and damage probability. The partial derivatives of the limit function of the damage level of UHPC-FHST components with respect to the four random variables are:

[0099]

[0100]

[0101] Therefore, the final component damage reliability and damage probability are: P = φ(-ψ), [β] is set to 0.95, 0.8 and 0.5 (under mild, moderate and severe injury conditions), respectively.

[0102] S4. Determine the reliability threshold based on the structural safety level to determine whether the component meets the performance requirements under impact, so as to achieve component performance design.

[0103] Specifically, step S4 includes:

[0104] According to the structural safety level, the reliability threshold is determined to determine whether the predicted component damage reliability is greater than the reliability threshold. If so, the component meets the performance requirements under impact. Otherwise, the reliability prediction is performed again by increasing the side length of the square section of the component, the thickness of the steel pipe, the yield strength of the steel pipe, and the compressive strength of the cubic concrete until the judgment conditions are met. The performance of the component meets the standard.

[0105] In this embodiment, a performance-based UHPC-FHST impact resistance design can be performed based on the reliability threshold, specifically:

[0106] By evaluating the correlation coefficient and analyzing the contribution of input parameters to the damage probability, we can directly reflect the degree of influence of various factors on damage, thus providing an effective reference for the formulation of component structure safety protection schemes. Random values ​​were taken within the database input parameter range to form 10,000 UHPC-FHST impact conditions, and the proposed method was used to calculate the moderate damage probability P; the Pearson correlation coefficient between each input parameter and P was calculated, and the absolute values ​​of each correlation coefficient were ranked as follows: Figure 8 As shown in the figure, the impact velocity V has a strong linear correlation with the damage probability P, with a correlation coefficient of 0.73, which is significantly higher than the other six input parameters; the correlation between P and column length L is very small, ranking seventh; for the component properties themselves, the square section side length B, steel yield strength f cu The influence of the thickness of the steel pipe t ranks in the top three, and can be considered as the starting point for anti-impact reinforcement.

[0107] At the same time, modern component structure design usually achieves performance design goals based on reliability. The reliability threshold [ψ] of component structure is often affected by the building environment, safety level, external load, etc. According to the sensitivity analysis results, when strengthening UHPC-FHST components for impact resistance, the cross-section side length B, steel yield strength f can be increased in turn. cu and the thickness of the steel pipe t; Therefore, based on the hybrid machine learning model SSA-Ann, the performance-based UHPC-FHST impact resistance design process proposed in this invention is as follows: 1) Select the impact parameters (M, V), component size parameters (L, B and t) and material parameters (f y 、f cu ) ; 2) Use the above formula to calculate the predicted residual bearing capacity coefficient β pre ; 3) The residual bearing capacity factor thresholds [β] for mild, moderate, and severe damage of UHPC-FHST components are taken as 0.95, 0.8, and 0.5, respectively, and the component damage reliability ψ under the impact load is calculated; 4) According to the standard "Uniform Standard for Reliability Design of Building Structures" (GB50068-2018), the structural safety level and the corresponding reliability threshold [ψ] are determined. If ψ> [ψ], the UHPC-FHST column meets the performance design requirements under the impact load; otherwise, the square section side length B and the steel yield strength f are increased. cu and the steel pipe thickness t, return to 1), and perform reliability prediction again until the judgment conditions are met, and the performance of the component meets the standards.

[0108] In summary, the reliability analysis method of high-strength steel tube ultra-high performance concrete components under axial impact proposed in the present invention firstly introduces the sparrow search method into the Ann neural network to establish a hybrid machine learning model, namely the SSA-Ann neural network model, to learn and train the residual bearing capacity coefficient of UHPC-FHST under axial impact, thereby improving the model stability and the prediction accuracy of the residual bearing capacity coefficient of the component; secondly, based on the training model, a residual bearing capacity coefficient prediction model of UHPC-FHST components after impact is established. Compared with theoretical research, the model formula does not require engineers to master profound physical principles (such as material nonlinearity, interaction between steel tube and concrete, second-order effects, etc.), which greatly facilitates designers. Then, the established database almost covers the material strength, structural geometry and impact load parameter range of high-performance steel tube ultra-high performance concrete columns, making the established model formula applicable to a wide range of applications. Finally, based on the application characteristics of UHPC-FHST components, the predicted residual bearing capacity is used to evaluate the damage level of UHPC-FHST components after impact. A calculation method for the reliability and damage probability of the components is further established. The random distribution of input parameter variables is taken into account based on the reliability. This ensures that in actual design, structural engineers can determine the reliability threshold for each damage level according to the structural safety level, and finally realize the design of the axial impact resistance of UHPC-FHST components.

[0109] Specific embodiments are used in the present invention to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core ideas. At the same time, for those skilled in the art, according to the ideas of the present invention, there may be changes in the specific implementation methods and application scopes. In summary, the contents of this specification should not be understood as limiting the present invention.

[0110] Those skilled in the art will appreciate that the embodiments described herein are intended to help readers understand the principles of the present invention, and it should be understood that the scope of protection of the present invention is not limited to such specific descriptions and embodiments. Those skilled in the art can make various other specific variations and combinations based on the technical teachings disclosed in the present invention without departing from the essence of the present invention, and such variations and combinations are still within the scope of protection of the present invention.

Claims

1. A reliability analysis method for axial impact of high-strength steel tube ultra-high performance concrete components, characterized by: The following steps are involved: S1. Collect axial impact sample sets of high-strength steel tube ultra-high performance concrete components. Each axial impact sample includes the impact load parameters, geometric parameters, and material parameters of the component. S2. Based on the axial impact sample set, the Ann neural network is trained and a residual bearing capacity coefficient prediction model is constructed to obtain the predicted residual bearing capacity coefficient; S3. Based on the predicted residual bearing capacity coefficient, the damage level of the component is obtained. By setting a threshold of the residual bearing capacity coefficient and combining the geometric parameters and material parameters of the component, a limit function of the component damage level is established to predict the component damage reliability and damage probability; S4. Determine the reliability threshold based on the structural safety level to determine whether the component meets the performance requirements under impact, so as to achieve component performance design.

2. The reliability analysis method for axial impact of high-strength steel tube ultra-high performance concrete components according to claim 1 is characterized in that: The impact load parameters of the component include impact mass and impact velocity; the geometric parameters include column length, square section side length, and steel pipe thickness; and the material parameters include steel pipe yield strength and cubic concrete compressive strength.

3. The reliability analysis method for axial impact of high-strength steel tube ultra-high performance concrete components according to claim 2 is characterized in that: Based on the axial impact sample set, the specific process of training the Ann neural network is as follows: Normalize the parameters of each axial impact sample in the axial impact sample set to obtain a normalized axial impact sample, namely: Among them, x ip represents the normalized value of the i-th parameter of the axial impact sample, x i Represents the original value of the i-th parameter of the axial impact sample, x i,min Indicates the minimum value of the i-th parameter of the axial impact sample, x i,max represents the maximum value of the i-th parameter of the axial impact sample; Based on the normalized axial impact samples, the Ann neural network is trained and a residual bearing capacity coefficient prediction model is constructed.

4. The reliability analysis method for axial impact of high-strength steel tube ultra-high performance concrete components according to claim 3 is characterized in that: Based on the normalized axial impact samples, the specific process of training the Ann neural network is as follows: Initialize the hyperparameters of the Ann neural network, including the number of hidden layers, the number of hidden layer neurons, different types of hidden layer transfer functions, and the weights and biases corresponding to each transfer function; Different types of hidden layer transfer functions and their corresponding weights and biases are used as inputs to the sparrow search method. By setting the population size, number of iterations, upper and lower limits of weights and biases, an iterative search is performed to obtain the optimal weights and biases. The transfer function corresponding to the optimal weight and bias is used as the optimal transfer function of the hidden layer; Based on the optimal transfer function of the hidden layer and its corresponding optimal weights and biases, the normalized axial impact samples are input into the Ann neural network for several training times, and finally a trained SSA-Ann neural network model is generated.

5. The reliability analysis method for axial impact of high-strength steel tube ultra-high performance concrete components according to claim 4 is characterized in that: The optimal transfer function of the hidden layer is a logarithmic transfer function.

6. The reliability analysis method for axial impact of high-strength steel tube ultra-high performance concrete components according to claim 5 is characterized in that: The process of constructing the residual bearing capacity coefficient prediction model is as follows: The normalized axial impact sample is input into the input layer of the trained SSA-Ann neural network model, and the first signal matrix is ​​output to the hidden layer. The first signal matrix received by the hidden layer is: α=xν+γ Where α represents the first signal matrix, x represents the normalized axial impact sample matrix, ν and γ represent the weight matrix and bias matrix between the input layer and the hidden layer, respectively; According to the optimal transfer function of the hidden layer, the second signal matrix is ​​output to the output layer, and the second signal matrix received by the output layer is: b=Logsig(α) Wherein, b represents the second signal matrix, and Logsig represents the logarithmic transfer function; The output signal of the final output layer is: λ=bw+θ Among them, λ represents the output signal, w and θ represent the weight matrix and bias matrix between the hidden layer and the output layer respectively; According to the output signal, a residual bearing capacity coefficient prediction model is constructed.

7. The reliability analysis method for axial impact of high-strength steel tube ultra-high performance concrete components according to claim 6 is characterized in that: The constructed residual bearing capacity coefficient prediction model is: Among them, β max , β min Respectively represent the maximum and minimum values ​​of the true value of the residual bearing capacity coefficient, β pre Represents the predicted residual bearing capacity coefficient.

8. The reliability analysis method for axial impact of high-strength steel tube ultra-high performance concrete components according to claim 7 is characterized in that: Step S3 specifically includes: S31. Based on the predicted residual bearing capacity coefficient, the component is classified into damage levels, including slight damage, moderate damage, and severe damage; S32. Obtaining coefficients of variation of geometric parameters and material parameters of the component, where the geometric parameters and material parameters of the component are normally distributed random variables; S33. Input the geometric parameters and material parameters of the component, and establish the limit function of the component damage level by setting the residual bearing capacity coefficient threshold, that is: G(B,t,f y ,f cu )=β pre -[b] Among them, G represents the limit function of component damage level, B represents the side length of the square section, t represents the thickness of the steel pipe, and f y Indicates the yield strength of the steel pipe, f vu represents the compressive strength of cubic concrete, [β] represents the threshold value of the residual bearing capacity coefficient; S34. Based on the limit function of component damage level, the first-order second moment method is used to calculate the component damage reliability, that is: Among them, ψ represents the component damage reliability, σ B , σ t 、 They represent the standard deviations of the side length of the square section, the thickness of the steel pipe, the yield strength of the steel pipe, and the compressive strength of the cubic concrete, respectively, and the standard deviations are obtained by multiplying the corresponding mean values ​​by the coefficient of variation; S35. Calculate the component damage probability based on the component damage reliability.

9. The reliability analysis method for axial impact of high-strength steel tube ultra-high performance concrete components according to claim 8 is characterized in that: The formula for calculating the component damage probability in step S34 is: P=φ(-ψ) Where P represents the component damage probability, and φ represents the cumulative function of the normal distribution function.

10. The reliability analysis method for axial impact of high-strength steel tube ultra-high performance concrete components according to claim 9, characterized in that: Step S4 specifically includes: According to the structural safety level, the reliability threshold is determined to determine whether the predicted component damage reliability is greater than the reliability threshold. If so, the component meets the performance requirements under impact. Otherwise, the reliability prediction is performed again by increasing the side length of the square section of the component, the thickness of the steel pipe, the yield strength of the steel pipe, and the compressive strength of the cubic concrete until the judgment conditions are met. The performance of the component meets the standard.

Images