CN112668078B - Method for identifying damage of rusted reinforced concrete beam after fire disaster - Google Patents

Abstract

A damage identification method for corroded reinforced concrete beams after fire, which relates to the technical field of safety risk assessment. The identification method is to identify damage indicators according to a combined algorithm of particle swarm optimization support vector machine method and residual error correction, including: step 1, using The ABAQUS finite element analysis software builds the damage model of the corroded reinforced concrete beam after the fire; step 2, calculates the structural combination parameter A according to the frequency and mode shape, and uses the structural combination parameter A as the input parameter of the combination algorithm; step 3, selects the indirect damage The index and the direct damage index are used as the output parameters of the combined algorithm; step 4, the radial basis kernel function is selected as the kernel function, and the original data of the low-dimensional space is mapped to the high-dimensional space; step 5, the mean square error and the best fit are selected. degree to measure the prediction accuracy; step 6, calculate according to the combination algorithm. The algorithm of the invention has obvious advantages for damage prediction, the calculation is more accurate, the time consumption is shorter, and the result reliability is higher.

Description

A post-fire damage identification method for corroded reinforced concrete beams

technical field

The invention relates to the technical field of safety risk assessment, in particular to a damage identification method for corroded reinforced concrete beams after fire.

Background technique

Fire is one of the causes of structural damage, collapse and casualties, as well as the main cause of direct and indirect economic losses. Structural damage identification and health monitoring technology can detect the existence and location of structural damage, and predict the remaining life of the structure. Therefore, it is of great engineering significance to establish a damage identification algorithm for structures to quickly and accurately identify the existence, location and extent of damage. Restricted by the complicated fire process, the expensive test equipment, and the damage effect of high temperature on the equipment, few tests have been carried out on the vibration characteristics of structures after a fire.

Traditional damage detection methods often rely too much on the experience of technicians, or cause different degrees of damage to the structure during the detection process. Most of the damage identification methods based on vibration test take a single frequency as the input index, ignoring the mode shape index which is more sensitive to damage. The damage identification method based on Support Vector Machine (SVM) has certain difficulties in the selection of the penalty parameter c and the kernel function parameter g, the convergence effect is not obvious, and the dimensional disaster is prone to occur. The damage identification algorithm based on particle swarm optimization support vector machine (PSO-SVM), in the process of parameter optimization of particle swarm optimization algorithm, when an example finds a current optimal position, other examples quickly move closer, and it is easy to fall into local optimum , the global optimal solution cannot be found.

At present, there are few studies on fire damage identification of corroded reinforced concrete beams, and most of them focus on damage identification at room temperature. Most of the damage identification methods based on SVM for vibration measurement do not have a good solution to the penalty parameters c and The selection of the kernel function parameter g leads to the disadvantages of large errors and time-consuming optimization when selecting parameters based on experience. However, in the process of damage identification and application of traditional algorithms (SVM, BP neural network, etc.), the model error, measurement error and environment The data errors caused by factors will make it insufficient in prediction accuracy and stability, and the fault tolerance and robustness are not ideal.

SUMMARY OF THE INVENTION

In order to solve the problems of the prior art, the present invention provides a damage identification method for corroded reinforced concrete beams after fire. The method superimposes a combined algorithm of residual error correction on the basis of the PSO-SVM algorithm, which can effectively solve the above problems, and the algorithm The advantages are obvious, the calculation is more accurate, the time-consuming is shorter, and the result is more reliable.

In order to solve the above-mentioned problems, the technical scheme of the present invention is:

A damage identification method for corroded reinforced concrete beams after fire, the identification method is to identify damage indicators according to a combined algorithm of particle swarm optimization support vector machine method and residual error correction, including: step 1, using ABAQUS finite element analysis software, constructing For the damage model of corroded reinforced concrete beams after fire, select Frequency analysis step, use Lanczos eigenvalue solver to perform modal analysis, and then obtain the modal parameters after fire damage; The residual bearing capacity is calculated, and the residual stiffness is calculated; step 2, according to the frequency and mode shape calculated in step 1 as the combination parameters, calculate the structural combination parameter A, and use the structural combination parameter A as the input parameter of the combination algorithm; step 3, Select the indirect damage index and the direct damage index as the output parameters of the combined algorithm; step 4, in the SVM of the combined algorithm, select the radial basis kernel function as the kernel function, and map the original data of the low-dimensional space into the high-dimensional space; Step 5. In the combined algorithm, the mean square error and the goodness of fit are selected as the accuracy evaluation indicators to measure the prediction accuracy of the damage index; Step 6, according to the combined algorithm of the particle swarm optimization support vector machine method and the residual error correction, The damage index of corroded reinforced concrete after fire is calculated.

Preferably, in the step 1, on the basis of modeling, based on the numerical simulation results of the temperature field, the literature values on the mass density, elastic modulus, and Poisson's ratio of the concrete and corroded steel bars after high temperature are selected as parameters, and the construction Damage model of corroded reinforced concrete beams after fire.

Preferably, in the step 1, the SPRING2 spring element is used to bond the reinforcement element and the concrete element, and the spring element stiffness of the reinforcement element node and the concrete element node in the horizontal direction is determined by the energy equivalent method.

Preferably, in the step 2, the calculation formula of the construction combination parameter A is:

A={FR ₁ ,FR ₂ ,...,FR _m ; MO ₁ ,MO ₂ ,...,MO _n }

where:

m: the order of the frequency

n: the order of the mode shape

FRi: the i-order frequency of the structure

为i阶模态下q个自由度的归一化振型向量，计算式为：

is the normalized mode shape vector of q degrees of freedom in the i-order mode, and the calculation formula is:

i阶模态下第j个自由度的振型分量

The mode shape component of the jth degree of freedom in the i-order mode

When constructing the combination parameter A, the following principles should be followed:

m≤4, n≤4.

Preferably, in the step 3, the calculation formulas of the indirect damage index and the direct damage index are respectively:

B={T,η _s }

C={α,β}

In the formula, B is the indirect damage index, C is the direct damage index, T is the fire exposure time, η _s is the corrosion rate, α is the bearing capacity reduction coefficient, and β is the stiffness reduction coefficient.

Preferably, in the step 5, the calculation formulas of the mean square error and the goodness of fit are respectively:

为预测值，n为测试样本个数。where MSE is the mean square error, R ² is the goodness of fit, y _i is the actual value,

is the predicted value, and n is the number of test samples.

Preferably, the step 6 includes the following specific algorithm steps:

(a) Establish a corroded beam structure model, the former third-order frequency mode shape is the input parameter, the direct damage index and the indirect damage index are the output parameters, and the training sample is constructed;

(b) Initialize the parameters of the particle swarm algorithm: population size, number of iterations, inertia weight, learning factor and the position and speed of each particle, all of which are randomly selected;

(c) Calculate the fitness value of each particle, and the fitness function takes the mean square error;

(d) Update the individual extreme value Pbest and the group extreme value Gbest of the particle according to the fitness value of the initial particle;

和

更新粒子的速度和位置，式中，ω为惯性权重；d为搜索空间的维数，d＝1,2,..,D；i＝1,2,…,n；k为当前送代次数；Vid为粒子的速度；c1和c2为非负的常数，称为加速因子；r1和r2为分布于[0,1]之间的随机数；(e) According to the formula

and

Update the speed and position of the particle, where ω is the inertia weight; d is the dimension of the search space, d=1,2,..,D; i=1,2,...,n; k is the current number of generations ; Vid is the velocity of the particle; c1 and c2 are non-negative constants called acceleration factors; r1 and r2 are random numbers distributed between [0,1];

(f) Update the individual extreme value and the group extreme value according to the particle fitness value in the new population;

(g) Judging whether the maximum number of iterations or the minimum error accuracy is satisfied, if satisfied, stop the iteration and output the optimal parameters; if not, go to step (c);

(h) applying the obtained optimal parameters to the SVM, and using the training set to train the SVM;

(i) Obtain the predicted value of the training set through training, and make the difference between the predicted value and the actual value to obtain the corresponding residual;

(j) Then train the same input and residual in the training set to obtain the residual predicted value;

(k) summing the residual prediction value and the prediction value of the initial algorithm to obtain a combined prediction value, and then constructing a combined prediction algorithm and a combined prediction model with the corresponding input and combined prediction value in the training set;

(1) Construct test samples using the measured modal parameters, and substitute them into the combined prediction model for verification;

(m) Judging whether the obtained output predicted value satisfies MSE≤0.05 and R ² ≥0.95, if so, output the result, if not, re-improve the model, and perform iterative optimization again.

The post-fire damage identification method for corroded reinforced concrete beams of the present invention has the following beneficial effects:

1. In the present invention, the frequency and mode shape sensitive to structural damage are used as the combined input parameters for post-fire damage identification, which is beneficial to improve the prediction accuracy.

2. Considering the damage assessment and reinforcement and repair after the fire, two sets of damage indicators are selected for the output parameters. The indirect damage indicators are the fire exposure time and the corrosion rate, and the direct damage indicators are the bearing capacity reduction coefficient and stiffness reduction coefficient. Among them, the indirect damage index is The loss index corresponds to the damage assessment after the fire, and the direct index corresponds to the reinforcement and repair after the fire.

3. The damage identification algorithm of PSO-SVM and residual correction solves the problem that the efficiency of manual parameter adjustment of SVM is too low during the use process, and alleviates the problem that PSO-SVM is easy to find local optimal parameters in the process of parameter optimization.

4. The damage identification algorithm of PSO-SVM and residual correction has strong robustness and fault tolerance, improves the convergence speed, greatly reduces the risk of adjusting parameters by experience, and has high prediction accuracy and stability.

Description of drawings

1, a schematic diagram of the stiffness of the spring unit in an embodiment of the present invention;

Fig. 2, the flow chart of the combined algorithm of particle swarm optimization support vector machine method and residual error correction of the present invention;

Figure 3. The beam reinforcement and thermocouple layout in the experimental example;

Detailed ways

The following describes the embodiments of the present invention in a step-by-step manner. This description is only a preferred embodiment of the present invention, and is not intended to limit the protection scope of the present invention. Anything within the spirit and principle of the present invention Any modifications, equivalent replacements and improvements made within the scope of the present invention should be included within the protection scope of the present invention.

In the description of the present invention, it should be noted that the orientation or positional relationship indicated by the terms "upper", "lower", "left", "right", "top", "bottom", "inside", "outside", etc. are based on those shown in the accompanying drawings. The orientation or positional relationship is only for describing the present invention and simplifying the description, rather than indicating or implying that the indicated device or element must have a specific orientation, as well as a specific orientation configuration and operation, and therefore should not be construed as a limitation of the present invention.

A damage identification method for corroded reinforced concrete beams after fire, the identification method is to identify damage indicators according to a combined algorithm of particle swarm optimization support vector machine method and residual error correction, including: step 1, using ABAQUS finite element analysis software, constructing For the damage model of corroded reinforced concrete beams after fire, select Frequency analysis step, use Lanczos eigenvalue solver to perform modal analysis, and then obtain the modal parameters after fire damage; The residual bearing capacity is calculated, and the residual stiffness is calculated; step 2, according to the frequency and mode shape calculated in step 1 as the combination parameters, calculate the structural combination parameter A, and use the structural combination parameter A as the input parameter of the combination algorithm; step 3, Select the indirect damage index and the direct damage index as the output parameters of the combined algorithm; step 4, in the SVM of the combined algorithm, select the radial basis kernel function as the kernel function, and map the original data of the low-dimensional space into the high-dimensional space; Step 5. In the combined algorithm, the mean square error and the goodness of fit are selected as the accuracy evaluation indicators to measure the prediction accuracy of the damage index; Step 6, according to the combined algorithm of the particle swarm optimization support vector machine method and the residual error correction, Calculate the damage index of corroded reinforced concrete after fire;

In the step 1, on the basis of modeling, based on the numerical simulation results of the temperature field, the literature values about the mass density, elastic modulus, and Poisson's ratio of concrete and corroded steel bars after high temperature are selected as parameters, and the post-fire corrosion is constructed. Damage model of reinforced concrete beam; as shown in Figure 1, according to the bond-slip constitutive relationship between corroded steel bars and concrete after fire obtained by the inventors in previous experiments (the curve in Figure 1, that is, the connection line of ABCDGF in the figure), such that S _OABCD = S _DEFG , where the slip value at point F of the residual segment is the slip value of the intersection point H of the extension line CG of the descending segment and the coordinate axis of the slip amount, and the slope corresponding to the line segment ODE is the spring element obtained by the energy equivalent method stiffness, the stiffness of the spring element in the vertical direction of the reinforcement element node and the concrete element node takes the maximum value;

In the step 1, the SPRING2 spring element is used for bonding between the reinforcement element and the concrete element, and the stiffness of the spring element in the horizontal direction between the reinforcement element node and the concrete element node is determined by the energy equivalent method;

In the described step 2, the calculation formula of the construction combination parameter A is:

A={FR ₁ ,FR ₂ ,...,FR _m ; MO ₁ ,MO ₂ ,...,MO _n }

where:

m: the order of the frequency

n: the order of the mode shape

FRi: the i-order frequency of the structure

为i阶模态下q个自由度的归一化振型向量，计算式为：

is the normalized mode shape vector of q degrees of freedom in the i-order mode, and the calculation formula is:

i阶模态下第j个自由度的振型分量

The mode shape component of the jth degree of freedom in the i-order mode

When constructing the combination parameter A, considering the actual engineering, the structural dynamic response test data has the characteristics of strong randomness, small amplitude, and easy to be polluted by noise. The following principles should be followed:

m≤4, n≤4; within this value range, the test result can be guaranteed to be valid;

In the step 3, the calculation formulas of the indirect damage index and the direct damage index are respectively:

B={T,η _s }

C={α,β}

In the formula, B is the indirect damage index, C is the direct damage index, T is the fire exposure time, η _s is the corrosion rate, α is the bearing capacity reduction coefficient, and β is the stiffness reduction coefficient; considering that the purpose of fire damage identification is To provide theoretical support for damage assessment and reinforcement and repair of structures after fire, the above indirect damage index and direct damage index are selected as output parameters in the algorithm;

In the step 5, the calculation formulas of the mean square error and the goodness of fit are respectively:

为预测值，n为测试样本个数；其中，MSE越趋近于0和R²越趋近于1代表算法的预测精度越高。where MSE is the mean square error, R ² is the goodness of fit, y _i is the actual value,

is the predicted value, n is the number of test samples; among them, the closer MSE is to 0 and the closer R ² is to 1, the higher the prediction accuracy of the algorithm is.

In the described step 6, the following specific algorithm steps are included, as shown in Fig. 2 for details:

(a) Establish a corroded beam structure model, the former third-order frequency mode shape is the input parameter, the direct damage index and the indirect damage index are the output parameters, and the training sample is constructed.

(b) Initialize the parameters of the particle swarm algorithm: population size, number of iterations, inertia weight, learning factor and the position and speed of each particle, all of which are randomly selected;

(c) Calculate the fitness value of each particle, and the fitness function takes the mean square error;

(d) Update the individual extreme value Pbest and the group extreme value Gbest of the particle according to the fitness value of the initial particle;

和

更新粒子的速度和位置，式中，ω为惯性权重；d为搜索空间的维数，d＝1,2,..,D；i＝1,2,…,n；k为当前送代次数；Vid为粒子的速度；c1和c2为非负的常数，称为加速因子；r1和r2为分布于[0,1]之间的随机数；(e) According to the formula

and

Update the speed and position of the particle, where ω is the inertia weight; d is the dimension of the search space, d=1,2,..,D; i=1,2,...,n; k is the current number of generations ; Vid is the velocity of the particle; c1 and c2 are non-negative constants called acceleration factors; r1 and r2 are random numbers distributed between [0,1];

(f) Update the individual extreme value and the group extreme value according to the particle fitness value in the new population;

(g) Judging whether the maximum number of iterations or the minimum error accuracy is satisfied, if satisfied, stop the iteration and output the optimal parameters; if not, go to step (c);

(h) applying the obtained optimal parameters to the SVM, and using the training set to train the SVM;

(i) Obtain the predicted value of the training set through training, and make the difference between the predicted value and the actual value to obtain the corresponding residual;

(j) Then train the same input and residual in the training set to obtain the residual predicted value;

(k) summing the residual prediction value and the prediction value of the initial algorithm to obtain a combined prediction value, and then constructing a combined prediction algorithm and a combined prediction model with the corresponding input and combined prediction value in the training set;

(1) Construct test samples using the measured modal parameters, and substitute them into the combined prediction model for verification;

(m) Judging whether the obtained output predicted value satisfies MSE≤0.05 and R ² ≥0.95, if so, output the result, if not, re-improve the model, and perform iterative optimization again.

Test example:

Design and manufacture scaled reinforced concrete beam specimens. The concrete beam is 3m long, of which the effective length is 2.4m, the section width is 150mm, the height is 300mm, and the thickness of the concrete protective layer is 20mm. C30 commercial concrete is used to cast the specimen. The steel bar adopts HRB400 steel bar with a diameter of 16mm. During the production process, one end of the tension steel bar is welded with a wire and drawn out for accelerated corrosion test by electrification. The erecting bars use HRB400 steel bars with a diameter of 12 mm, and the stirrups use HPB300 steel bars with a diameter of 8 mm, with a spacing of 100 mm. The detailed experimental design parameters are shown in Table 1, of which the B1 beam is the comparison beam, and the beam reinforcement and thermocouple layout are shown in the figure. 3.

Table 1 Experimental design parameters

The dynamic test was carried out on the corroded reinforced concrete beams before and after the fire, and the modal measurement points were selected as the six equal points between the supports.

The damage identification results after the fire are shown in Tables 2 and 3.

Table 2 Identification results of indirect damage index after fire

y_i是实际值，

为预测值。Statement: T is the equivalent detonation time, and T _p is the predicted value of the equivalent detonation time. The equivalent detonation time is the time corresponding to the standard heating curve with the same area under the actual heating curve. η _s is the corrosion rate, and η _s,p is the predicted value of the corrosion rate.

y _i is the actual value,

is the predicted value.

Table 3 Identification results of direct damage index after fire

Claims (6)

1. A method for identifying damage of a rusted reinforced concrete beam after a fire disaster is characterized by comprising the following steps: the identification method is used for identifying the damage index according to a particle swarm optimization support vector machine method and a residual correction combined algorithm, and comprises the following steps: step 1, establishing a damage model of a corroded reinforced concrete beam after fire by using ABAQUS finite element analysis software, selecting a Frequency analysis step, and performing modal analysis by using a Lanczos eigenvalue solver to further obtain modal parameters after fire damage; meanwhile, a static force general analysis step is selected to obtain the residual bearing capacity of the corroded reinforced concrete beam after fire disaster, and the residual rigidity is calculated; step 2, calculating a construction combination parameter A according to the frequency and the vibration mode calculated in the step 1 as combination parameters, and using the construction combination parameter A as an input parameter of a combination algorithm; step 3, selecting the indirect damage index and the direct damage index as output parameters of a combined algorithm; step 4, selecting a radial basis kernel function as a kernel function in the SVM of the combined algorithm, and mapping the original data of the low-dimensional space to the high-dimensional space; step 5, in the combined algorithm, selecting the mean square error and the goodness of fit as precision evaluation indexes to measure the prediction precision of the damage index; step 6, calculating damage indexes of the rusted reinforced concrete after the fire disaster according to a particle swarm optimization support vector machine method and a combined algorithm of residual correction;

the step 6 comprises the following specific algorithm steps:

(a) establishing a corrosion beam structure model, taking the former three-order frequency vibration mode as an input parameter, taking a direct damage index and an indirect damage index as output parameters, and constructing a training sample;

(b) initializing parameters of a particle swarm algorithm: the population scale, the iteration times, the inertia weight, the learning factor, the position and the speed of each particle and the like are randomly valued;

(c) calculating the fitness value of each particle, and taking the mean square error as a fitness function;

(d) updating individual extreme value Pbest and group extreme value Gbest of the particles according to the fitness value of the initial particles;

(e) according to the formula

And

updating the speed and the position of the particle, wherein omega is an inertia weight; d is the dimension of the search space, D1, 2. 1,2, …, n; k is the current generation times; vid is the velocity of the particles; c1 and c2 are non-negative constants called acceleration factors; r1 and r2 are distributed in [0,1 ]]A random number in between;

(f) updating individual extreme values and population extreme values according to the particle fitness values in the new population;

(g) judging whether the maximum iteration times or the minimum error precision is met, if so, stopping iteration and outputting an optimal parameter; if not, turning to the step (c);

(h) applying the obtained optimal parameters to the SVM, and training the SVM by using a training set;

(i) obtaining a predicted value of a training set through training, and subtracting the predicted value from an actual value to obtain a corresponding residual error;

(j) then training the same input and residual in the training set to obtain a residual predicted value;

(k) summing the residual prediction value and the prediction value of the initial algorithm to obtain a combined prediction value, and constructing a combined prediction algorithm by using the corresponding input and combined prediction values in the training set to construct a combined prediction model;

(l) Constructing a test sample by using the actual measurement modal parameters, and substituting the test sample into the combined prediction model for verification;

(m) judging whether the obtained output prediction value meets MSE (mean Square error) of less than or equal to 0.05 and R²And (4) being more than or equal to 0.95, if the result is satisfied, outputting the result, and if the result is not satisfied, re-perfecting the model and performing iterative optimization again.

2. The method for identifying the damage of the rusted reinforced concrete beam after the fire disaster as claimed in claim 1, is characterized in that: in the step 1, on the basis of modeling, literature values of mass density, elastic modulus and Poisson's ratio of the concrete and the corroded steel bar after high temperature are selected as parameters based on a temperature field numerical simulation result, and a corroded reinforced concrete beam damage model after fire is constructed.

3. The method for identifying the damage of the rusted reinforced concrete beam after the fire disaster as claimed in claim 1, is characterized in that: in the step 1, the steel bar units and the concrete units are bonded by SPRING units SPRING2, and the SPRING unit stiffness of the steel bar unit nodes and the concrete unit nodes in the horizontal direction is determined by an energy equivalence method.

4. The method for identifying the damage of the rusted reinforced concrete beam after the fire disaster as claimed in claim 1, is characterized in that: in the step 2, the calculation formula for constructing the combination parameter a is as follows:

A＝{FR₁,FR₂,…,FR_m；MO₁,MO₂,…,MO_n}

in the formula:

m: order of frequency

n: order of mode shape

F, FRi: i order frequency of structure

The normalized mode shape vector of q degrees of freedom under the i-order mode is calculated by the following formula:

mode shape component of jth freedom degree under i-order mode

When constructing the combination parameter a, the following principle is followed:

m≤4，n≤4。

5. the method for identifying the damage of the rusted reinforced concrete beam after the fire disaster as claimed in claim 1, is characterized in that: in step 3, the calculation formulas of the indirect damage index and the direct damage index are respectively as follows:

B＝{T,η_s}

C＝{α,β}

wherein B is an indirect damage index, C is a direct damage index, T is an ignition time, eta_sFor corrosion rate, α is the reduction coefficient of bearing capacity and β is the reduction coefficient of stiffness.

6. The method for identifying the damage of the rusted reinforced concrete beam after the fire disaster as claimed in claim 1, is characterized in that: in step 5, the calculation formulas of the mean square error and the goodness of fit are respectively as follows:

where MSE is the mean square error, R²Is goodness of fit, y_iIs the actual value of the,

is a predicted value, and n is the number of test samples.

Images