A bridge vulnerability assessment method based on debris flow dynamic process

Abstract

The present invention discloses a bridge vulnerability assessment method based on the dynamic process of debris flow, which relates to the technical field of debris flow damage monitoring. The method comprises: obtaining the quantitative relationship between the debris flow intensity and the pier top displacement through artificial neural network model training and evaluation, statistically analyzing the pier top displacement of multiple samples obtained by the artificial neural network model to obtain the pier top displacement probability distribution function, and combining the vulnerability index classification standard to obtain the failure probability of the bridge pier. The vulnerability curve is drawn by using different debris flow intensities as independent variables in turn to perform Monte Carlo sampling to draw the vulnerability curve. The method can timely grasp the probability of different degrees of damage to the bridge under the impact of debris flows of different intensities, so as to timely carry out disaster prevention and disaster relief work.

Description

Technical Field

[0001] The invention relates to the technical field of debris flow monitoring, and in particular to a bridge vulnerability assessment method based on a debris flow dynamic process. Background Art

[0002] With the continuous improvement of structural response research and field disaster history data, structural response research provides a quantitative means for debris flow vulnerability, and field disaster history data supports the vulnerability curve. For single structures, researchers selected a certain structural dynamic parameter as vulnerability through structural response research and established a functional relationship between debris flow intensity and structural response parameters. Zhang et al. (2016) conducted 9 sets of 1:1 indoor model experiments of iron balls of different heights and masses hitting brick-concrete walls. According to the brick-concrete wall failure criterion, the fuzzy comprehensive evaluation method was used to quantify the wall cracks and maximum dynamic displacement as vulnerability, and the vulnerability curves corresponding to momentum, maximum impact force and maximum impact bending moment were obtained. Luo et al. (2020) used LS-DYNA (ALE) to study the failure mode of steel-concrete frame buildings under the action of water-rock flow, debris flow, flood and earth flow. The article considered the fluid as a single-phase fluid and established four impact conditions with increasing solid concentration by increasing the solid concentration. The intrinsic physical properties of different fluids on building damage were compared. Some scholars believe that material strength, on-site construction and later structural durability cannot be completely consistent with the design, and the uncertainty of the model needs to be considered when conducting vulnerability analysis. The two uncertainty models proposed by Eidsvig et al. (2014) are widely used to deal with the uncertainty of debris flow intensity, evaluation model uncertainty and building damage determination uncertainty: 1. Establish the confidence band interval of the vulnerability function. 2. Calculate the exceedance probability under different damage states. The calculation of failure probability can be expressed as the conditional probability that the structure reaches or exceeds the component damage under the action of a debris flow of a given intensity, and can also be expressed as the probability that the structural demand under the action of a debris flow exceeds the structure's ability to resist debris flow disasters. Jakob et al. (2012) used momentum flux as the debris flow intensity index for 68 buildings with detailed records of debris flow depth and velocity, and classified the economic losses into four categories. Using statistical methods, a probability matrix of different damage states of buildings under different debris flow intensities was established. Prieto et al. (2018) used the American HAZUS tsunami model to establish a relationship function between debris flow impact force and the lateral resistance of buildings. By grading the lateral resistance of buildings into four types of damage states, the debris flow impact force was calculated using momentum flux, thereby establishing a exceedance probability function curve between debris flow momentum flux and damage state. Liang and Xiong (2019) established a numerical model of 150 groups of debris flow impact bridge piers by using different combinations of debris flow depth and velocity, calculated the pier top limit displacement based on static analysis, and used the nonlinear finite element method to solve the pier top displacement under each working condition. The pier top displacement / limit displacement was used as the bridge damage index, and a vulnerability curve between debris flow momentum flux and exceedance probability was established.(Material uncertainty, aging properties) Yan et al. (2020) proposed a vulnerability analysis method for bridge piers under the action of debris flows based on nonlinear structural analysis. This method takes into account the uncertainties of pier materials such as concrete strength and steel bar strength, and uses Poisson square wave process and pulse process to characterize the impact of debris flows. An uncertainty model is established through Monte Carlo sampling and nonlinear finite element analysis is performed. The vulnerability curves of low-viscosity and high-viscosity debris flow depth, flow velocity and pier failure probability are established.

[0003] However, these technologies currently have the following problems: although regional and single-channel models and methods have been initially formed, the dynamic performance of debris flows on bridges has not been studied in depth, so the assignment of model weights and parameters is still subjective. Therefore, current research rarely establishes vulnerability assessment models based on mechanical mechanisms. Summary of the invention

[0004] The present invention aims to provide a bridge vulnerability assessment method based on the dynamic process of debris flow, so as to accurately grasp the probability of different degrees of damage to the bridge under the impact of debris flow of different intensities, and to carry out disaster prevention and mitigation work in a timely manner.

[0005] The present invention provides a bridge vulnerability assessment method based on debris flow dynamic process, comprising:

[0006] The dynamic process of debris flow impacting the bridge pier is simulated by using the fluid-solid coupling simulation method, the dynamic response value of the bridge under different debris flow intensity variables is obtained, and the dynamic response value of the bridge is used as the pier top displacement;

[0007] Using the bridge dynamic response values ​​under the different debris flow intensity variables, an artificial neural network model is constructed and trained to obtain a trained artificial neural network model;

[0008] Calculating a bending moment-curvature curve of a pier section, and determining a vulnerability index classification using the bending moment-curvature curve of the pier section;

[0009] The trained artificial neural network model and the vulnerability index classification are used to calculate the failure probability of piers under different damage states, and the pier failure probability is used to perform a vulnerability assessment of the bridge project.

[0010] Preferably, the debris flow intensity variables include: flow velocity v, flow depth h, rock radius r and debris flow impact angle α.

[0011] Preferably, in the fluid-solid coupling simulation method, the fluid model adopts the finite volume method based on CFX for numerical analysis, the mud model adopts the VOF two-phase flow, the fluid viscosity model adopts the k-eps model, the mud adopts the Bingham body model, the stone model adopts the spherical stone model, the concrete model adopts the RHT model, and the steel bar model adopts the bilinear isotropic elastoplastic hardening model.

[0012] Preferably, the use of the bridge dynamic response values ​​under the different debris flow intensity variables to construct and train an artificial neural network model to obtain a trained artificial neural network model includes:

[0013] Constructing an artificial neural network model, and dividing the different debris flow intensity variables and the bridge dynamic response values ​​under the different debris flow intensity variables into a training set of the artificial neural network model and a prediction set of the artificial neural network model;

[0014] The artificial neural network model is trained using the training set to obtain a trained artificial neural network model, and the accuracy of the trained artificial neural network model is evaluated using the training set and the prediction set to confirm the accuracy of the trained artificial neural network model.

[0015] Preferably, the calculating the moment-curvature curve of the pier cross section and determining the vulnerability index classification using the moment-curvature curve of the pier cross section comprises:

[0016] By calculating the resistance of the pier concrete structure, the moment-curvature relationship of the pier section is established;

[0017] The moment-curvature relationship of the pier cross section is fitted into a bilinear curve, and the vulnerability index classification is determined using the bilinear curve.

[0018] Preferably, the calculating the failure probability of the pier under different damage states by using the trained artificial neural network model and the vulnerability index classification comprises:

[0019] Determine the probability distribution function of each debris flow intensity, and select n debris flow intensity groups by Monte Carlo sampling method;

[0020] Inputting the n debris flow intensity groups into the trained artificial neural network model respectively to obtain n pier top displacements, and obtaining a probability distribution function of the pier top displacements by counting the n pier top displacements;

[0021] The n pier top displacements are classified by using the vulnerability index classification to obtain the number of pier top displacements under different damage states, and the failure probability of the pier under different damage states is calculated according to the number of pier top displacements under different damage states and the probability distribution function of the pier top displacements.

[0022] Preferably, the probability distribution function for determining the intensity of each debris flow comprises:

[0023] When the flow velocity and the flow depth satisfy the Weibull distribution, determining the probability density and the probability distribution function of the flow velocity and the flow depth;

[0024] When the stone radius satisfies the Gumbel distribution, determining the probability density and probability distribution function of the stone radius;

[0025] When the impact angle interval is 0 degrees to 90 degrees, the probability density and probability distribution function of the impact angle are determined.

[0026] Preferably, the use of the bridge pier failure probability to perform bridge engineering vulnerability assessment includes:

[0027] A fragility curve is drawn according to the probability distribution function of the pier top displacement under the different debris flow intensity variables, and the fragility curve is used to evaluate the fragility of the bridge project.

[0028] In summary, the beneficial effects of the present invention are as follows: the dynamic response process of debris flow impacting bridge piers is taken into account, a bridge vulnerability risk assessment method based on probability analysis is constructed, and it is helpful to conduct quantitative risk assessment of mountain bridges. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 It is a flow chart of a bridge vulnerability assessment method based on debris flow dynamic process provided by an embodiment of the present invention;

[0030] Figure 2 is a diagram of a debris flow impacting a bridge pier model provided by an embodiment of the present invention;

[0031] Figure 3 is a diagram of a geometric model of a debris flow impacting a bridge pier provided by an embodiment of the present invention;

[0032] Figure 4 is an artificial neural network learning model evaluation diagram provided by an embodiment of the present invention;

[0033] Figure 5 is a bending moment-curvature curve of a pier cross section provided by an embodiment of the present invention;

[0034] Figure 6 is a flow chart of Monte Carlo solution to bridge pier failure probability provided by an embodiment of the present invention;

[0035] Figure 7 is a probability distribution curve diagram of debris flow intensity parameters provided by an embodiment of the present invention;

[0036] Figure 8 is a probability distribution diagram of pier top displacement provided by an embodiment of the present invention;

[0037] Fig. 9 is a pier vulnerability curve diagram provided by an embodiment of the present invention;

[0038] Fig.10 The present invention provides a flow chart of a bridge vulnerability assessment method based on debris flow dynamic process. DETAILED DESCRIPTION

[0039] The preferred embodiments of the present invention are described in detail below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described below are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0040] Embodiment 1:

[0041] Fig.10 is a flow chart of a bridge vulnerability assessment method based on debris flow dynamic process provided by an embodiment of the present invention, such as Fig.10 As shown, including:

[0042] Step S101: using a fluid-solid coupling simulation method to simulate the dynamic process of debris flow impacting a bridge pier, obtaining a bridge dynamic response value under different debris flow intensity variables, and using the bridge dynamic response value as the pier top displacement;

[0043] Step S102: constructing and training an artificial neural network model using the bridge dynamic response values ​​under the different debris flow intensity variables to obtain a trained artificial neural network model;

[0044] Step S103: calculating the bending moment-curvature curve of the pier cross section, and determining the vulnerability index classification using the bending moment-curvature curve of the pier cross section;

[0045] Step S104: using the trained artificial neural network model and the vulnerability index classification, calculating the failure probability of the piers under different damage states, and using the pier failure probability to perform bridge engineering vulnerability assessment.

[0046] The debris flow intensity variables include: flow velocity v, flow depth h, rock radius r and debris flow impact angle α.

[0047] Specifically, in the fluid-solid coupling simulation method, the fluid model adopts the finite volume method based on CFX for numerical analysis, the mud model adopts the VOF two-phase flow, the fluid viscosity model adopts the k-eps model, the mud adopts the Bingham body model, the stone model adopts the spherical stone model, the concrete model adopts the RHT model, and the steel bar model adopts the bilinear isotropic elastoplastic hardening model.

[0048] Among them, the use of the bridge dynamic response values ​​under the different debris flow intensity variables to construct and train the artificial neural network model to obtain the trained artificial neural network model includes: constructing an artificial neural network model, and dividing the different debris flow intensity variables and the bridge dynamic response values ​​under the different debris flow intensity variables into a training set of the artificial neural network model and a prediction set of the artificial neural network model; using the training set to train the artificial neural network model to obtain the trained artificial neural network model, and using the training set and the prediction set to evaluate the accuracy of the trained artificial neural network model to confirm the accuracy of the trained artificial neural network model.

[0049] Furthermore, the calculation of the moment-curvature curve of the pier cross section and the determination of the vulnerability index classification using the moment-curvature curve of the pier cross section include: establishing the moment-curvature relationship of the pier cross section by calculating the resistance of the pier concrete structure; fitting the moment-curvature relationship of the pier cross section into a bilinear curve, and determining the vulnerability index classification using the bilinear curve.

[0050] The method of calculating the failure probability of the piers under different damage states by using the trained artificial neural network model and the vulnerability index classification includes: determining the probability distribution function of each debris flow intensity, and taking n debris flow intensity groups by Monte Carlo sampling method; inputting the n debris flow intensity groups into the trained artificial neural network model respectively to obtain n pier top displacements, and obtaining the probability distribution function of the pier top displacements by counting the n pier top displacements; classifying the n pier top displacements by using the vulnerability index classification to obtain the number of pier top displacements under different damage states, and calculating the failure probability of the piers under different damage states according to the number of pier top displacements under different damage states and the probability distribution function of the pier top displacements.

[0051] Specifically, the probability distribution function for determining the intensity of each debris flow includes: when the flow velocity and flow depth satisfy the Weibull distribution, determining the probability density and probability distribution function of the flow velocity and the flow depth; when the stone radius satisfies the Gumbel distribution, determining the probability density and probability distribution function of the stone radius; when the impact angle range is 0 degrees to 90 degrees, determining the probability density and probability distribution function of the impact angle.

[0052] Wherein, the use of the pier failure probability to perform bridge engineering vulnerability assessment includes: drawing a vulnerability curve according to the probability distribution function of the pier top displacement under the different debris flow intensity variables, and using the vulnerability curve to perform bridge engineering vulnerability assessment.

[0053] Embodiment 2:

[0054] The main design process of this embodiment is as follows Figure 1 As shown in the figure, by simulating the dynamic process of debris flow impact, the data set of debris flow intensity and bridge pier dynamic response is obtained, and multivariate parameter regression fitting is performed through machine learning to obtain the functional relationship between debris flow intensity and bridge pier dynamic response. Combined with the vulnerability index classification standard, the failure probability of the bridge pier is calculated by the Monte Carlo sampling method, and the vulnerability curve is drawn; specifically, it includes:

[0055] Establish a debris flow impact bridge pier model Figure 2 As shown. The fluid domain is a cuboid, 11.5m long, 8m wide, and 7m high. The bottom 4m is the debris flow domain, and the top 3m is the air domain. A double-column pier is placed in the center of the basin, and a spherical stone with a radius of 1m is placed directly in front of the pier. The double-column pier is 11.6m high and 0.9m in diameter. The center distance between the two piers is 3.5m. The tie beam section is 0.7m high and 0.4m wide. The foundation is a rigid foundation. The concrete is C35, the protective layer thickness is 40mm, the longitudinal reinforcement is HRB400, 18C32; the stirrups are ordinary stirrups HRB335, B14@200, and the pier top load is 4500kN. The pier geometry model is as follows Figure 3 shown.

[0056] The debris flow intensity velocity, flow depth, stone radius and impact angle are taken as independent variables, and the influence of bridge pier deformation on the flow field is ignored. The dynamic process of debris flow impacting bridge pier is simulated by using unidirectional fluid-solid coupling simulation, and the displacement of the bridge pier top is obtained as the dependent variable to obtain the debris flow intensity and the dynamic response U of the bridge pier. s Datasets;

[0057] Among them, the fluid simulation adopts the finite volume method based on CFX for numerical analysis, and the density of the debris flow is 1750kg / m3. The fluid adopts the Bingham body model, the dynamic viscosity coefficient is 0.7Pa·s, and the yield stress is 16 Pa. The mud model adopts the VOF two-phase flow, the main components are air and mud. When the volume of mud in the unit reaches 50%, the unit is occupied by mud, and the velocity and other properties of the unit are mud properties. The fluid viscosity model adopts the k-eps model, and the mud adopts the Bingham body model. Since the impact force of the debris flow is mainly concentrated on the faucet, that is, the impact process is related to time, transient analysis is adopted, the time step is 0.005s, and the calculation is 500 steps, that is, 2.5s. In the initialization calculation, the fluid domain and the air domain are filled with air, the debris flow inlet velocity is 10m / s, and the flow depth is 4m; the air inlet is set as the pressure inlet at the boundary of the flow domain and the pressure difference is set to 0, the outlets are all pressure outlets, and the top of the air domain is a connected opening with a pressure of 1 atmosphere.

[0058] The stone model adopts a spherical stone model, and the stone is regarded as a sphere. The parameters of the stone such as elastic modulus and volume density are shown in Table 1.

[0059]

[0060] The concrete model adopts the RHT model, which is widely used to simulate the dynamic loading of brittle materials such as rock, concrete and ceramic impact processes. The RHT constitutive model is a combined model of plasticity and shear damage, in which the deviatoric stress in the material is limited by the generalized failure surface, and its failure surface is mainly manifested in the elastic limit surface, failure limit surface and residual strength limit surface. Each limit surface can be expressed as initial yield strength, peak yield strength and post-peak residual strength. The RHT constitutive equation divides the compression stage into three stages: linear elastic stage, plastic transition stage, and fully dense material stage. The porous material state equation P-ɑ model is used in the plastic compression stage, and the polynomial state equation describing the dense material is used in the dense stage.

[0061] The steel bar model uses a bilinear isotropic elastic-plastic hardening model, which can be used for large deformation analysis. The initial slope of the curve is the material modulus E, and the second-stage slope is the tangent elastic modulus Et. Therefore, to define the steel bar model, you only need to input the steel bar yield strength σ0, the initial elastic modulus E, and the tangent elastic modulus Et. The tangent elastic modulus cannot be greater than the initial elastic modulus.

[0062] The velocity v, k and epsilon are used as residual indicators to calculate the residual to determine whether the calculation has converged. -3 , the calculation converges.

[0063] The above simulation randomly selects values ​​of four independent variables, namely velocity, flow depth, rock radius and impact angle, in a reasonable range, and performs 170 sets of numerical simulations to obtain the data set of pier top displacement, which is divided into 126 training sets and 44 test sets. The training is carried out through the artificial neural network model control model. The impact angle, flow depth, rock radius and flow velocity are used as independent variables, and the pier top displacement is used as the dependent variable. All data in the model generation process and prediction process are set to be positive values. The mean squared error (MSE) function and adaptive moment estimation (Adam) optimization algorithm are used to evaluate the training model error (Heaton, 2011). The neural network training continuously adjusts the weights in the hidden layer according to the error between the calculated predicted value and the actual value to achieve the effect of optimizing the model. When the calculated error drops to an acceptable threshold, the training is terminated to obtain the ideal model. The input set of the training model is the debris flow impact angle, flow depth, rock radius and flow velocity, and the output set is the pier top displacement, so as to achieve the data fitting of debris flow intensity and pier dynamic response.

[0064] After obtaining the machine learning model based on the training sample parameters, the model error needs to be analyzed. The present invention uses the comparison of training samples and test samples to evaluate the model accuracy (Table 2).l is the actual pier top displacement in the training sample, U lc is the pier top displacement calculated by substituting the strength parameters in the training sample into the model; U t is the actual pier top displacement in the test sample, U tc The pier top displacement is calculated by substituting the strength parameters of the test sample into the model.

[0065]

[0066] The original data is used as the horizontal axis and the model calculation data is used as the vertical axis. Figure 4 As shown, the training sample set is (Ul, Ulc) and the test sample set is (Ut, Utc). When the model calculation data is closer to the real data, the sample point is closer to the straight line y=x, indicating that the model is more accurate. The error graph shows that both the training sample and the test sample are closer to the straight line, that is, the real data and the calculated data are closer. The training sample and the test sample are linearly fitted, and the slope of the straight line is 0.854, which is closer to the ideal straight line y=x, proving that the training model has a good accuracy effect.

[0067] Considering that the damage of debris flow to bridge piers is mainly bending damage, and the displacement ductility ratio is based on the moment-curvature curve of the steel-concrete section when it is bent, and is derived by selecting different damage stages of the control section. This study uses the displacement ductility ratio as the vulnerability index μ. Before classifying the structural damage index, when establishing the moment-curvature relationship of the pier section, the moment-curvature curve of the pier section is calculated by using XTRACT and fitted into a bilinear curve, such as Figure 5 As shown; the initial elastic stage is the line connecting the origin and the first yield point of the tensile steel bar, and the plastic stage is the line connecting the equivalent yield point and the limit curvature point. The equivalent yield point is calculated by the equality of the bilinear first straight line area and the original curve area.

[0068] The vulnerability index classification standard uses the displacement ductility ratio to classify the pier damage. The main classification methods and related descriptions are shown in the table.

[0069]

[0070] in , , , They are first yield displacement ductility ratio, equivalent yield displacement ductility ratio, and concrete compressive strain The displacement ductility ratio and the maximum displacement ductility ratio at the time of the displacement ductility ratio are shown in the following formula.

[0071] (9)

[0072] (10)

[0073] (11)

[0074] (12)

[0075] (13)

[0076] In the formula , , are the first yielding time of steel bar, equivalent yield point and concrete compressive strain respectively. The corresponding top displacement (m) is is the displacement of the pier top under the lateral load (m). Concrete compressive strain The corresponding top displacement Displacement by elastic deformation and plastic hinge displacement Composition, each parameter is calculated as follows.

[0077] (14)

[0078] (15)

[0079] (16)

[0080] (17)

[0081] (18)

[0082] In the formula According to the load combination, take 1 / 3, plastic hinge length Calculate according to formula (19). , , are the first yield curvature, equivalent yield curvature, and concrete compressive strain in cross-section analysis. The corresponding curvature.

[0083] (19)

[0084] Where fy is the design value of the tensile strength of the steel bar (MPa), ds is the diameter of the steel bar (m), h is the distance from the dangerous section to the inflection point (m), and D is the diameter of the pier (m)

[0085] The calculation method of the failure probability Pf of the bridge pier is shown in formula (20), where UR is the vulnerability index classification, as shown in Table 3, and US is the displacement response of the pier top under the action of debris flow, which is calculated according to the above artificial neural network model as follows:

[0086] (20)

[0087] (twenty one)

[0088] There are two types of uncertainties in the vulnerability analysis of bridges to debris flows that affect the assessment results: accidental uncertainty and epistemic uncertainty. In the present invention, accidental uncertainty refers to the randomness of debris flows, and epistemic uncertainty refers to the uncertainty of bridge structure parameters. The present invention assumes that the bridge design parameters are known, that is, there is no epistemic uncertainty and only the accidental uncertainty caused by the debris flow intensity parameters needs to be considered. Since the debris flow intensity parameter is the independent variable of the pier top displacement function, and the probability distribution of the debris flow intensity parameter needs to be considered when obtaining it, it is difficult to directly solve the failure probability function of the bridge pier, so Monte Carlo sampling is used to solve the failure probability.

[0089] Determine the probability distribution function of each debris flow intensity parameter, including the following:

[0090] The flow velocity v and flow depth h satisfy the Weibull distribution, and their probability density and probability function are shown in formula (2) and formula (3).

[0091] (2)

[0092] (3)

[0093] Among them, λ and k are the scale parameter and shape parameter respectively, and their specific values ​​are calculated based on the relevant terrain and flow.

[0094] Assuming that the stone radius satisfies the Gumbel distribution, the probability density and probability function of the stone radius R are shown in formulas (4)-(6):

[0095] (4)

[0096] (5)

[0097] (6)

[0098] Among them, μ is the location parameter, β is the scale parameter, and is the Euler constant with a value of 0.577.

[0099] Among them, μ is the location parameter, β is the scale parameter, and is the Euler constant with a value of 0.577.

[0100] The impact angle range is from 0 to 90 degrees. The cumulative distribution function and probability density are shown in formulas (7) and (8).

[0101] (7)

[0102] (8)

[0103] Among them, μ is the mean, σ is the standard deviation, and the value is 0.3;

[0104] The failure probability of the bridge pier is solved by Monte Carlo calculation. The specific process is as follows: Figure 6 shown.

[0105] According to the probability distribution function of debris flow intensity parameters, n debris flow intensity groups X1...Xn, are selected, where the i-th debris flow intensity Xi=[vi, hi, ri,αi];

[0106] Substitute Xi into the artificial neural network model to get the pier top displacement Ui, substitute n groups of Xi into the artificial neural network model to get US composed of n pier top displacements, and then the probability distribution function of the pier top displacement can be obtained by counting Us;

[0107] The pier top displacement Us is classified according to the vulnerability index classification standard, and the number of pier top displacements under different damage states C=[c1,c2,c3,c4] is counted respectively;

[0108] Calculate the failure probability of the bridge pier under different damage states:

[0109] Pf=[c1 / n, c2 / n, c3 / n, c4 / n] (22)

[0110] When the number of samples n is small, the sample randomness is large and the accuracy is poor. As the number of independent variable groups increases, the calculation of failure probability becomes more and more accurate. When the number of independent variable groups is infinite, it can be equivalent to solving the failure probability. To ensure the calculation accuracy, the number of independent variable groups is 108.

[0111] Before drawing the vulnerability curve, the probability distribution function of the pier top displacement under the action of uncertain debris flow intensity parameters is first obtained. The specific method is: according to the probability distribution function, n debris flow intensity parameters are taken, substituted into the machine learning model, n pier top displacements are obtained, and then the distribution function of the pier top displacement is fitted according to the probability distribution.

[0112] Figure 7is the sampling value of debris flow intensity parameters, where the impact angle satisfies the normal distribution, the mean is π / 4, and the coefficient of variation is 0.38. The mean values ​​of flow velocity and flow depth are 6m / s and 4m respectively. Both are sampled according to Weibull distribution, and the coefficient of variation (COV) is 0.3. The stone radius is sampled according to Gumbel distribution, with a mean of 0.7m and a coefficient of variation of 0.3. To ensure the calculation accuracy, a total of 108 parameters are sampled. Substitute the debris flow intensity value into the artificial neural network model to calculate an array of 108 pier top displacements, count the histogram frequency of the pier top displacements and perform probability density function fitting, as shown in Figure 8 shown.

[0113] (twenty three)

[0114] The displacement of the pier top is fitted using 80 types of probability distribution functions, and the first three types of distribution functions that fit well are selected, namely gamma distribution, Gumbel distribution and normal distribution, with goodness of fit R2 of 0.94, 0.84 and 0.73 respectively. It can be seen that the displacement of the pier top under the action of debris flow satisfies the gamma distribution, as shown in formula (23).

[0115] The Monte Carlo sampling is used as independent variables to draw the vulnerability curve, as shown in the following figure: Fig. 9 shown.

[0116] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A bridge vulnerability assessment method based on debris flow dynamic process, characterized in that: The following steps are involved: The dynamic process of debris flow impacting the bridge pier is simulated by using the fluid-solid coupling simulation method, the dynamic response value of the bridge under different debris flow intensity variables is obtained, and the dynamic response value of the bridge is used as the pier top displacement; Using the bridge dynamic response values ​​under the different debris flow intensity variables, an artificial neural network model is constructed and trained to obtain a trained artificial neural network model; Calculating a bending moment-curvature curve of a pier section, and determining a vulnerability index classification using the bending moment-curvature curve of the pier section; The trained artificial neural network model and the vulnerability index classification are used to calculate the failure probability of piers under different damage states, and the pier failure probability is used to perform a vulnerability assessment of the bridge project.

2. The method according to claim 1, characterized in that The debris flow intensity variables include: flow velocity v, flow depth h, rock radius r and debris flow impact angle α.

3. The method according to claim 1, characterized in that In the fluid-solid coupling simulation method, the fluid model adopts the finite volume method based on CFX for numerical analysis, the mud model adopts the VOF two-phase flow, the fluid viscosity model adopts the k-eps model, the mud adopts the Bingham body model, the stone model adopts the spherical stone model, the concrete model adopts the RHT model, and the steel bar model adopts the bilinear isotropic elastoplastic hardening model.

4. The method according to claim 2, characterized in that: The artificial neural network model is constructed and trained by using the bridge dynamic response values ​​under the different debris flow intensity variables to obtain the trained artificial neural network model, which includes: Constructing an artificial neural network model, and dividing the different debris flow intensity variables and the bridge dynamic response values ​​under the different debris flow intensity variables into a training set of the artificial neural network model and a prediction set of the artificial neural network model; The artificial neural network model is trained using the training set to obtain a trained artificial neural network model, and the accuracy of the trained artificial neural network model is evaluated using the training set and the prediction set to confirm the accuracy of the trained artificial neural network model.

5. The method according to claim 4, characterized in that The calculating of the moment-curvature curve of the pier section and determining the vulnerability index classification using the moment-curvature curve of the pier section includes: By calculating the resistance of the pier concrete structure, the moment-curvature relationship of the pier section is established; The moment-curvature relationship of the pier cross section is fitted into a bilinear curve, and the vulnerability index classification is determined using the bilinear curve.

6. The method according to claim 5, characterized in that The method of calculating the failure probability of bridge piers under different damage states by using the trained artificial neural network model and the vulnerability index classification includes: Determine the probability distribution function of each debris flow intensity, and select n debris flow intensity groups by Monte Carlo sampling method; Inputting the n debris flow intensity groups into the trained artificial neural network model respectively to obtain n pier top displacements, and obtaining a probability distribution function of the pier top displacements by counting the n pier top displacements; The n pier top displacements are classified by using the vulnerability index classification to obtain the number of pier top displacements under different damage states, and the failure probability of the pier under different damage states is calculated according to the number of pier top displacements under different damage states and the probability distribution function of the pier top displacements.

7. The method according to claim 6, characterized in that The probability distribution function for determining the intensity of each debris flow includes: When the flow velocity and the flow depth satisfy the Weibull distribution, determining the probability density and the probability distribution function of the flow velocity and the flow depth; When the stone radius satisfies the Gumbel distribution, determining the probability density and probability distribution function of the stone radius; When the impact angle interval is 0 degrees to 90 degrees, the probability density and probability distribution function of the impact angle are determined.

8. The method according to claim 7, characterized in that The method of using the bridge pier failure probability to carry out bridge engineering vulnerability assessment includes: A fragility curve is drawn according to the probability distribution function of the pier top displacement under the different debris flow intensity variables, and the fragility curve is used to evaluate the fragility of the bridge project.

Images