WO2021051332A1 - Bridge seismic damage monitoring method based on wavelet neural network and support vector machine - Google Patents
Bridge seismic damage monitoring method based on wavelet neural network and support vector machine Download PDFInfo
- Publication number
- WO2021051332A1 WO2021051332A1 PCT/CN2019/106578 CN2019106578W WO2021051332A1 WO 2021051332 A1 WO2021051332 A1 WO 2021051332A1 CN 2019106578 W CN2019106578 W CN 2019106578W WO 2021051332 A1 WO2021051332 A1 WO 2021051332A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- vector machine
- support vector
- wavelet neural
- neural network
- wavelet
- Prior art date
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01D—MEASURING NOT SPECIALLY ADAPTED FOR A SPECIFIC VARIABLE; ARRANGEMENTS FOR MEASURING TWO OR MORE VARIABLES NOT COVERED IN A SINGLE OTHER SUBCLASS; TARIFF METERING APPARATUS; MEASURING OR TESTING NOT OTHERWISE PROVIDED FOR
- G01D21/00—Measuring or testing not otherwise provided for
- G01D21/02—Measuring two or more variables by means not covered by a single other subclass
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
Definitions
- the invention relates to the technical field of bridge damage monitoring, in particular to a bridge earthquake damage monitoring method based on wavelet neural network and support vector machine.
- structural damage monitoring methods mainly include model-based methods and data-driven methods.
- the model-based method has a large amount of calculation, which is difficult in practical applications and cannot realize online damage monitoring.
- the data-driven damage monitoring method has high calculation accuracy and can realize online damage monitoring, so it has great value in earthquake damage monitoring.
- commonly used data-driven methods include artificial intelligence methods, support vector machine methods, particle swarm methods, etc. Considering the complexity of the bridge structure environment, it is of practical significance to adopt a combined multiple data-driven seismic damage monitoring method.
- the purpose of the present invention is to provide a bridge seismic damage monitoring method based on wavelet neural network and support vector machine.
- the functional relationship between bearing damage and main girder displacement time history is established based on the dynamic balance equation, and the bearing damage sample is established through the finite element model.
- Database select the RBF kernel function to simplify the solution of the nonlinear approximation problem, use wavelet neural network to train the support vector machine network parameters, and then use the wavelet support vector machine method to achieve the approximate solution of the above functional relationship. After checking its recognition accuracy, you can It is applied to bridge earthquake damage monitoring to solve the problems raised in the background art.
- Bridge seismic damage monitoring method based on wavelet neural network and support vector machine includes the following steps:
- Step 1 Build a sample database
- a finite element model of the full bridge is established, and the different damaged samples and undamaged samples of the proposed bearings are substituted into the finite element model, and the maximum displacement of the main girder under different seismic load conditions is simulated.
- Value, the spring stiffness index of the support and the corresponding maximum displacement of the main beam are used as the sample database of the wavelet support vector machine;
- Step 2 Determine the range of wavelet neural network parameter values
- Step 3 Self-training of wavelet neural network to obtain test errors of different wavelet neural networks
- n groups of wavelet neural network parameters to form n wavelet neural networks; use the training sample library A described in the first step to train the wavelet neural network , And then use the verification sample library B for testing to obtain the test errors of different wavelet neural networks;
- Step 4 Support vector machine network training to get the optimal wavelet neural network parameters
- the test errors corresponding to the n groups of wavelet neural network parameters are obtained; then the optimal wavelet neural network parameters can be obtained based on the support vector machine network training;
- Step 5 Damage monitoring based on wavelet neural network and support vector machine
- the actual seismic load can be obtained The change of the spring stiffness of the lower support, so as to identify the damage state of the support.
- m groups of samples are selected based on the D optimal design method, and the sample database is divided into training sample database A and verification sample database B.
- the seismic wave time history data described in the first step can be included in the latest seismic wave records at any time, and the training sample bank A is m-10 groups, and the verification sample bank B is 10 groups.
- the beneficial effects of the present invention are: the present invention is based on the wavelet neural network and the support vector machine bridge seismic damage monitoring method, based on the dynamic balance equation to establish the time history function relationship between the support damage and the main beam displacement,
- the RBF kernel function is selected to simplify the solution of the nonlinear approximation problem, and the wavelet neural network is used to train the support vector machine network parameters, and then the wavelet support vector machine method is used to achieve the approximate solution of the above function relationship. After checking its recognition accuracy, it can be applied Earthquake damage monitoring of real bridge bearings.
- the invention adopts wavelet neural network to provide a more effective means for solving the support vector machine network parameters, the bearing earthquake damage prediction result is good, and it has broad application prospects in the field of structural earthquake engineering in the future.
- Fig. 1 is a flow chart of earthquake damage monitoring of bridge bearings according to the present invention
- Figure 2 is a diagram of the wavelet neural network model of the present invention.
- Fig. 3 is a time-history response diagram of the displacement of the bridge 1 under the action of an earthquake according to the present invention
- Fig. 4 is a time-history response diagram of the displacement of the bridge 2 under the action of an earthquake of the present invention.
- the method for monitoring bridge seismic damage based on wavelet neural network and support vector machine includes the following steps:
- the first step establish a sample database
- the finite element model of the full bridge is established, and the different damaged samples and undamaged samples of the proposed bearings are substituted into the finite element model together.
- the wavelet neural network model is shown in Figure 2, and the simulation results are obtained under different seismic load conditions.
- the main beam displacement time history data under the action, the spring stiffness index of the support and the corresponding main beam displacement time history data are used as the sample database of the wavelet support vector machine, and m groups of samples are selected based on the D optimal design method, and the sample database Divided into training sample bank A and parameter determination sample bank B;
- the above-mentioned support spring stiffness is based on the environmental vibration test of the bridge in the undamaged state, and the finite element model is modified to obtain the initial value of the support spring stiffness damage. According to the investigation data of the bridge seismic damage, the change range of the damage parameter is determined. The invention controls it within (0.05-1) times the initial value.
- Step 2 Determine the range of support vector machine network parameter values
- Step 3 Self-training of wavelet neural network to obtain test errors of different wavelet neural networks
- the training sample bank A of trains the wavelet neural network, and then uses the verification sample bank B to test to obtain the test errors of different wavelet neural networks.
- the results are shown in Table 2.
- the fourth step support vector machine network training to obtain the optimal wavelet neural network parameters
- Step 5 Damage monitoring based on wavelet neural network and support vector machine
- the difference between the present invention and the prior art is: in the existing bridge structure damage monitoring, the optimization principle of the support vector machine network is usually selected to solve the function relationship between the stiffness and acceleration of the structure, and the support vector machine regression network Training accuracy is closely related to the selection of its internal parameter values, and support vector machine networks often use empirical formulas or trial calculations in actual engineering applications. This is random and not universal, and it is easier to cause local optimization of the network. And mutations.
- the bridge bearing damage monitoring method based on wavelet neural network and support vector machine provided by the present invention establishes the functional relationship between bearing damage and main girder displacement time history based on the dynamic balance equation, and selects the RBF kernel function to simplify the nonlinear approximation problem.
- the present invention is based on the wavelet neural network and support vector machine bridge seismic damage monitoring method, based on the dynamic balance equation to establish the time-history function relationship between the support damage and the main girder displacement, and the RBF kernel function is selected to simplify the nonlinear approximation problem.
- the RBF kernel function is selected to simplify the nonlinear approximation problem.
- the invention adopts wavelet neural network to provide a more effective means for solving the support vector machine network parameters, the bridge earthquake damage prediction result is good, and it has broad application prospects in the field of structural earthquake engineering in the future.
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
- Bridges Or Land Bridges (AREA)
Abstract
A bridge seismic damage monitoring method based on a wavelet neural network and a support vector machine. First, a function relationship between bearing damage and a main girder displacement time history is established on the basis of a dynamic equilibrium equation; a bridge seismic damage sample database is established by means of a finite element model; a solution of an RBF kernel function simplified nonlinear approximation problem is selected; network parameters of the support vector machine are trained by using the wavelet neural network, and thus an approximate solution of the function relationship is achieved by using a wavelet support vector machine method, and after the identification accuracy is checked, the method can be applied to seismic damage monitoring of a real bridge bearing. The bridge seismic damage monitoring method based on the wavelet neural network and the support vector machine is more universal with respect to seeking of the network parameters of the support vector machine by means of an empirical formula or trial calculation, provides a more effective means for bridge seismic damage monitoring, has a good prediction result in bridge seismic damage, and has a wide application prospect in the field of structural seismic engineering in the future.
Description
本发明涉及桥梁损伤监控技术领域，特别涉及基于小波神经网络和支持向量机桥梁地震损伤监控方法。The invention relates to the technical field of bridge damage monitoring, in particular to a bridge earthquake damage monitoring method based on wavelet neural network and support vector machine.
中国有数以万计的桥梁和高架桥组成的一个复杂的高速公路系统，这些桥梁结构随着时间的推移、会因为一系列的自然灾害，如地震、火灾、风暴、长期腐蚀以及疲劳的作用，损伤会不断的累积。早期识别地震引起的损害是特别重要的，以确定该结构在地震后是否可以继续使用。China has a complex highway system composed of tens of thousands of bridges and viaducts. Over time, these bridge structures will be damaged by a series of natural disasters, such as earthquakes, fires, storms, long-term corrosion and fatigue. Will continue to accumulate. Early recognition of the damage caused by the earthquake is particularly important to determine whether the structure can continue to be used after the earthquake.
目前，结构损伤监控方法主要有基于模型的方法和基于数据驱动的方法。其中，基于模型的方法计算量较大，在实际应用中较为困难，且无法实现在线损伤监控。而基于数据驱动的损伤监控方法，计算精度较高，且能实现在线损伤监控，故而其在地震损伤监控方面具有较大价值。在工程领域，应用较为普遍的数据驱动方法有人工智能方法、支持向量机方法、粒子群方法等，考虑到桥梁结构环境的复杂性，采用联合多种数据驱动的地震损伤监控方法具有现实意义。At present, structural damage monitoring methods mainly include model-based methods and data-driven methods. Among them, the model-based method has a large amount of calculation, which is difficult in practical applications and cannot realize online damage monitoring. The data-driven damage monitoring method has high calculation accuracy and can realize online damage monitoring, so it has great value in earthquake damage monitoring. In the engineering field, commonly used data-driven methods include artificial intelligence methods, support vector machine methods, particle swarm methods, etc. Considering the complexity of the bridge structure environment, it is of practical significance to adopt a combined multiple data-driven seismic damage monitoring method.
发明内容Summary of the invention
本发明的目的在于提供基于小波神经网络和支持向量机桥梁地震损伤监控方法，首先基于动力学平衡方程建立了支座损伤与主梁位移时程间函数关系，通过有限元模型建立支座损伤样本数据库，选取RBF核函数简化非线性逼近问题的求解，用小波神经网络对支持向量机网络参数进行训练，进而利用小波支持向量机方法实现了上述函数关系的近似求解，检验其识别精度后，可以应用于桥梁地震损伤监控，以解决上述背景技术中提出的问题。The purpose of the present invention is to provide a bridge seismic damage monitoring method based on wavelet neural network and support vector machine. Firstly, the functional relationship between bearing damage and main girder displacement time history is established based on the dynamic balance equation, and the bearing damage sample is established through the finite element model. Database, select the RBF kernel function to simplify the solution of the nonlinear approximation problem, use wavelet neural network to train the support vector machine network parameters, and then use the wavelet support vector machine method to achieve the approximate solution of the above functional relationship. After checking its recognition accuracy, you can It is applied to bridge earthquake damage monitoring to solve the problems raised in the background art.
为实现上述目的，本发明提供如下技术方案：In order to achieve the above objectives, the present invention provides the following technical solutions:
基于小波神经网络和支持向量机桥梁地震损伤监控方法，包括以下步骤：Bridge seismic damage monitoring method based on wavelet neural network and support vector machine includes the following steps:
第一步：建立样本数据库Step 1: Build a sample database
根据被测试桥梁的设计资料，建立全桥的有限元模型，将拟定的支座不同损伤样本与未损伤样本一同代入有限元模型，分别模拟得到在不同地震荷载工况作用下的主梁位移最大值，将支座的弹簧刚度指标和相应的主梁位移最大值作为小波支持向量机的样本数据库；According to the design data of the tested bridge, a finite element model of the full bridge is established, and the different damaged samples and undamaged samples of the proposed bearings are substituted into the finite element model, and the maximum displacement of the main girder under different seismic load conditions is simulated. Value, the spring stiffness index of the support and the corresponding maximum displacement of the main beam are used as the sample database of the wavelet support vector machine;
第二步：确定小波神经网络参数值的范围Step 2: Determine the range of wavelet neural network parameter values
调查大量国内外小波神经网络研究现状，拟定其网络各参数值的范围，包括优误差控制参数ε、惩罚因子C和核函数参数σ；Investigate the current research status of a large number of wavelet neural networks at home and abroad, and draw up the range of each parameter value of the network, including the optimal error control parameter ε, the penalty factor C and the kernel function parameter σ;
第三步：小波神经网络自训练，得到不同小波神经网络的测试误差Step 3: Self-training of wavelet neural network to obtain test errors of different wavelet neural networks
在上述第二步的所述的小波神经网络参数值范围内，选择n组小波神经网络参数，形成n个小波神经网络；利用上述第一步所述的训练样本库A对小波神经网络进行训练，再利用验证样本库B进行测试，得到不同小波神经网络的测试误差；Within the parameter value range of the wavelet neural network in the second step above, select n groups of wavelet neural network parameters to form n wavelet neural networks; use the training sample library A described in the first step to train the wavelet neural network , And then use the verification sample library B for testing to obtain the test errors of different wavelet neural networks;
第四步：支持向量机网络训练，得到最优小波神经网络参数Step 4: Support vector machine network training to get the optimal wavelet neural network parameters
基于上述第三步计算结果，得到了n组小波神经网络参数分别对应的测试误差；进而基于支持向量机网络训练可以得到最优的小波神经网络参数；Based on the calculation results of the third step above, the test errors corresponding to the n groups of wavelet neural network parameters are obtained; then the optimal wavelet neural network parameters can be obtained based on the support vector machine network training;
第五步：基于小波神经网络和支持向量机的损伤监控Step 5: Damage monitoring based on wavelet neural network and support vector machine
将地震作用下实测得到的位移最大值作为损伤后的代表值，并结合损伤前支座弹簧刚度与主梁位移计算值，代入上述已训练好的小波支持向量机网络，可以得到实际地震荷载作用下支座弹簧刚度的变化，从而对支座损伤状态进行识别。Taking the maximum displacement measured under the earthquake action as the representative value after the damage, combined with the calculated value of the spring stiffness of the support before the damage and the displacement of the main girder, and substituting the trained wavelet support vector machine network into the above-mentioned trained wavelet support vector machine network, the actual seismic load can be obtained The change of the spring stiffness of the lower support, so as to identify the damage state of the support.
进一步地，第一步中基于D最优设计方法选取m组样本，并将样本数据库分成训练样本库A和验证样本库B。Further, in the first step, m groups of samples are selected based on the D optimal design method, and the sample database is divided into training sample database A and verification sample database B.
进一步地，第一步所述地震波时程数据可以随时计入最新的地震波记录，且训练样本库A为m-10组，验证样本库B为10组。Further, the seismic wave time history data described in the first step can be included in the latest seismic wave records at any time, and the training sample bank A is m-10 groups, and the verification sample bank B is 10 groups.
进一步地，第二步支持向量机网络训练过程中，核函数选择RBF核函数，即k(x
i,x)＝exp(-|x
i-x|
2/2σ
2)。
Further, in the second step of the support vector machine network training process, the kernel function selects the RBF kernel function, that is, k(x i ,x)=exp(-|x i -x| 2 /2σ 2 ).
与现有技术相比，本发明的有益效果是：本发明基于小波神经网络和支持向量机桥梁地震损伤监控方法，基于动力学平衡方程建立了支座损伤与主梁位移时程间函数关系，选取RBF核函数简化非线性逼近问题的求解，用小波神经网络对支持向量机网络参数进行训练，进而利用小波支持向量机方法实现了上述函数关系的近似求解，检验其识别精度后，可以应用于实桥支座的地震损伤监控。本发明采用小波神经网络，为求解支持向量机网络参数提供了更加有效的手段，支座地震损伤预测结果良好，在未来的结构地震工程领域具有广泛的应用前景。Compared with the prior art, the beneficial effects of the present invention are: the present invention is based on the wavelet neural network and the support vector machine bridge seismic damage monitoring method, based on the dynamic balance equation to establish the time history function relationship between the support damage and the main beam displacement, The RBF kernel function is selected to simplify the solution of the nonlinear approximation problem, and the wavelet neural network is used to train the support vector machine network parameters, and then the wavelet support vector machine method is used to achieve the approximate solution of the above function relationship. After checking its recognition accuracy, it can be applied Earthquake damage monitoring of real bridge bearings. The invention adopts wavelet neural network to provide a more effective means for solving the support vector machine network parameters, the bearing earthquake damage prediction result is good, and it has broad application prospects in the field of structural earthquake engineering in the future.
图1为本发明的桥梁支座地震损伤监控流程图；Fig. 1 is a flow chart of earthquake damage monitoring of bridge bearings according to the present invention;
图2为本发明的小波神经网络模型图；Figure 2 is a diagram of the wavelet neural network model of the present invention;
图3为本发明的地震作用下桥梁1的位移时程响应图；Fig. 3 is a time-history response diagram of the displacement of the bridge 1 under the action of an earthquake according to the present invention;
图4为本发明的地震作用下桥梁2的位移时程响应图。Fig. 4 is a time-history response diagram of the displacement of the bridge 2 under the action of an earthquake of the present invention.
下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本发明一部分实施例，而不是全部的实施例。基于本发明中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都属于本发明保护的范围。The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative work shall fall within the protection scope of the present invention.
参阅图1，基于小波神经网络和支持向量机桥梁地震损伤监控方法，包括以下步骤：Refer to Figure 1. The method for monitoring bridge seismic damage based on wavelet neural network and support vector machine includes the following steps:
第一步：建立样本数据库；The first step: establish a sample database;
根据被测试桥梁的设计资料，建立全桥的有限元模型，将拟定的支座不 同损伤样本与未损伤样本一同代入有限元模型，小波神经网络模型如图2，模拟得到在不同地震荷载工况作用下的主梁位移时程数据，将支座的弹簧刚度指标和相应的主梁位移时程数据作为小波支持向量机的样本数据库，基于D最优设计方法选取m组样本，并将样本数据库分成训练样本库A和参数确定样本库B；According to the design data of the tested bridge, the finite element model of the full bridge is established, and the different damaged samples and undamaged samples of the proposed bearings are substituted into the finite element model together. The wavelet neural network model is shown in Figure 2, and the simulation results are obtained under different seismic load conditions. The main beam displacement time history data under the action, the spring stiffness index of the support and the corresponding main beam displacement time history data are used as the sample database of the wavelet support vector machine, and m groups of samples are selected based on the D optimal design method, and the sample database Divided into training sample bank A and parameter determination sample bank B;
上述支座弹簧刚度，是基于未损伤状态的桥梁环境振动试验，展开有限元模型修正，可得到支座弹簧刚度损伤的初始值，根据桥梁地震损伤的调查资料，确定损伤参数的变化范围，本发明将其控制在(0.05-1)倍初始值内。The above-mentioned support spring stiffness is based on the environmental vibration test of the bridge in the undamaged state, and the finite element model is modified to obtain the initial value of the support spring stiffness damage. According to the investigation data of the bridge seismic damage, the change range of the damage parameter is determined. The invention controls it within (0.05-1) times the initial value.
地震波时程数据可以随时计入最新的地震波记录，且训练样本库A为m-10组，参数确定样本库为10组，其中m＝100，如表1所示。The seismic wave time history data can be included in the latest seismic wave records at any time, and the training sample bank A is m-10 groups, and the parameter determination sample bank is 10 groups, where m=100, as shown in Table 1.
表1Table 1
第二步：确定支持向量机网络参数值的范围Step 2: Determine the range of support vector machine network parameter values
调查大量国内外支持向量机研究现状，针对桥梁地震损伤调查资料，拟定其网络各参数值的范围，包括优误差控制参数ε、惩罚因子C和核函数参数σ。Investigate a large number of domestic and foreign support vector machines research status, based on the bridge seismic damage investigation data, draw up the range of the network parameters, including the optimal error control parameter ε, the penalty factor C and the kernel function parameter σ.
第三步：小波神经网络自训练，得到不同小波神经网络的测试误差Step 3: Self-training of wavelet neural network to obtain test errors of different wavelet neural networks
在上述第二步的所述的小波神经网络参数值范围内，选择n组小波神经网络参数，形成n个小波神经网络，其中n＝100，如表2所示；利用上述第一步所述的训练样本库A对小波神经网络进行训练，再利用验证样本库B进行测试，得到不同小波神经网络的测试误差，结果如表2所示。Within the parameter value range of the wavelet neural network in the second step above, select n groups of wavelet neural network parameters to form n wavelet neural networks, where n=100, as shown in Table 2; use the above-mentioned first step The training sample bank A of, trains the wavelet neural network, and then uses the verification sample bank B to test to obtain the test errors of different wavelet neural networks. The results are shown in Table 2.
表2Table 2
| 组数Number of groups | εε |
C | σσ | 误差error | |
| 11 | 0.00010.0001 | 0.20.2 | 1.51.5 | 0.0150.015 | |
| 22 | 0.0050.005 | 0.310.31 | 2.12.1 | 0.0140.014 | |
| 33 | 0.000080.00008 | 0.150.15 | 3.63.6 | 0.0240.024 | |
| …… | …… | …… | …… | …… | |
| 100100 | 0.000020.00002 | 0.240.24 | 10.410.4 | 0.0010.001 |
第四步：支持向量机网络训练，得到最优小波神经网络参数；The fourth step: support vector machine network training to obtain the optimal wavelet neural network parameters;
基于上述第三步计算结果，得到了n组小波神经网络参数分别对应的测试误差；进而基于支持向量机网络训练可以得到最优的小波神经网络参数；误差控制参数ε＝0.0001、惩罚因子C＝0.21和核函数参数σ＝9.1。Based on the calculation results of the third step above, the test errors corresponding to the n groups of wavelet neural network parameters are obtained; then the optimal wavelet neural network parameters can be obtained based on the support vector machine network training; the error control parameter ε = 0.0001, the penalty factor C = 0.21 and the kernel function parameter σ=9.1.
第五步：基于小波神经网络和支持向量机的损伤监控；Step 5: Damage monitoring based on wavelet neural network and support vector machine;
将地震作用下实测得到的位移最大值作为损伤后的代表值，如图3-4所示，代入上述已训练好的小波支持向量机网络，不同的位移最大值所对应的桥梁支座刚度。其中，图3所示，位移最大值为18.5mm，对应的支座刚度为0.95；图4所示，位移最大值为39.1mm，对应的支座刚度为0.38，由此可知：图3所对应的桥梁支座刚度损伤5％；图4所对应桥梁支座损伤为62％，从而对支座损伤状态进行识别。Take the maximum displacement measured under the earthquake action as the representative value after damage, as shown in Figure 3-4, substituting the trained wavelet support vector machine network into the above-mentioned trained wavelet support vector machine network, and the stiffness of the bridge support corresponding to the different maximum displacements. Among them, as shown in Figure 3, the maximum displacement is 18.5mm, and the corresponding support stiffness is 0.95; as shown in Figure 4, the maximum displacement is 39.1mm, and the corresponding support stiffness is 0.38. It can be seen that: The stiffness damage of the bridge support is 5%; Figure 4 corresponds to the bridge support damage of 62%, so that the damage state of the support can be identified.
本发明与现有技术的区别在于：在现有的桥梁结构损伤监控中，通常选择采用支持向量机网络的最优化原理对结构的刚度与加速度的函数关系进行求解，且支持向量机回归网络的训练精度与其内部参数值的选取密切相关，而支持向量机网络在实际工程应用中往往采用经验公式或试算寻求，这样随机性较大，不具备普适性，较易造成网络的局部最优和突变。本发明提供的基于小波神经网络和支持向量机的桥梁支座损伤监控方法，基于动力学平衡方程建立了支座损伤与主梁位移时程间函数关系，选取RBF核函数简化非线性逼近问题的求解，用小波神经网络对支持向量机网络参数进行训练，进而 利用小波支持向量机方法实现了上述函数关系的近似求解，检验其识别精度后，可以应用于桥梁地震损伤监控。本发明采用小波神经网络，为求解支持向量机网络参数提供了更加有效的手段，桥梁地震损伤预测结果良好，在未来的结构地震工程领域具有广泛的应用前景。The difference between the present invention and the prior art is: in the existing bridge structure damage monitoring, the optimization principle of the support vector machine network is usually selected to solve the function relationship between the stiffness and acceleration of the structure, and the support vector machine regression network Training accuracy is closely related to the selection of its internal parameter values, and support vector machine networks often use empirical formulas or trial calculations in actual engineering applications. This is random and not universal, and it is easier to cause local optimization of the network. And mutations. The bridge bearing damage monitoring method based on wavelet neural network and support vector machine provided by the present invention establishes the functional relationship between bearing damage and main girder displacement time history based on the dynamic balance equation, and selects the RBF kernel function to simplify the nonlinear approximation problem. Solve, use wavelet neural network to train the support vector machine network parameters, and then use wavelet support vector machine method to achieve the approximate solution of the above-mentioned functional relationship. After checking its recognition accuracy, it can be applied to bridge earthquake damage monitoring. The invention adopts wavelet neural network to provide a more effective means for solving the support vector machine network parameters, the bridge earthquake damage prediction result is good, and it has broad application prospects in the field of structural earthquake engineering in the future.
综上所述：本发明基于小波神经网络和支持向量机桥梁地震损伤监控方法，基于动力学平衡方程建立了支座损伤与主梁位移时程间函数关系，选取RBF核函数简化非线性逼近问题的求解，用小波神经网络对支持向量机网络参数进行训练，进而利用小波支持向量机方法实现了上述函数关系的近似求解，检验其识别精度后，可以应用于桥梁地震损伤监控。本发明采用小波神经网络，为求解支持向量机网络参数提供了更加有效的手段，桥梁地震损伤预测结果良好，在未来的结构地震工程领域具有广泛的应用前景。To sum up: the present invention is based on the wavelet neural network and support vector machine bridge seismic damage monitoring method, based on the dynamic balance equation to establish the time-history function relationship between the support damage and the main girder displacement, and the RBF kernel function is selected to simplify the nonlinear approximation problem. To solve the problem, use wavelet neural network to train the support vector machine network parameters, and then use wavelet support vector machine method to achieve the approximate solution of the above functional relationship. After checking its recognition accuracy, it can be applied to bridge earthquake damage monitoring. The invention adopts wavelet neural network to provide a more effective means for solving the support vector machine network parameters, the bridge earthquake damage prediction result is good, and it has broad application prospects in the field of structural earthquake engineering in the future.
以上所述，仅为本发明较佳的具体实施方式，但本发明的保护范围并不局限于此，任何熟悉本技术领域的技术人员在本发明披露的技术范围内，根据本发明的技术方案及其发明构思加以等同替换或改变，都应涵盖在本发明的保护范围之内。The above are only the preferred specific embodiments of the present invention, but the protection scope of the present invention is not limited to this. Anyone familiar with the technical field within the technical scope disclosed in the present invention, according to the technical solution of the present invention Equivalent replacements or changes to its inventive concept should all fall within the protection scope of the present invention.
Claims (4)
- 基于小波神经网络和支持向量机桥梁地震损伤监控方法，其特征在于，包括以下步骤：The method for monitoring bridge seismic damage based on wavelet neural network and support vector machine is characterized in that it includes the following steps:第一步：建立样本数据库Step 1: Build a sample database根据被测试桥梁的设计资料，建立全桥的有限元模型，将拟定的支座不同损伤样本与未损伤样本一同代入有限元模型，分别模拟得到在不同地震荷载工况作用下的主梁位移最大值，将支座的弹簧刚度指标和相应的主梁位移最大值作为小波支持向量机的样本数据库；According to the design data of the tested bridge, a finite element model of the full bridge is established, and the different damaged samples and undamaged samples of the proposed bearings are substituted into the finite element model, and the maximum displacement of the main girder under different seismic load conditions is simulated. Value, the spring stiffness index of the support and the corresponding maximum displacement of the main beam are used as the sample database of the wavelet support vector machine;第二步：确定小波神经网络参数值的范围Step 2: Determine the range of wavelet neural network parameter values调查大量国内外小波神经网络研究现状，拟定其网络各参数值的范围，包括优误差控制参数ε、惩罚因子C和核函数参数σ；Investigate the current research status of a large number of wavelet neural networks at home and abroad, and draw up the range of each parameter value of the network, including the optimal error control parameter ε, the penalty factor C and the kernel function parameter σ;第三步：小波神经网络自训练，得到不同小波神经网络的测试误差Step 3: Self-training of wavelet neural network to obtain test errors of different wavelet neural networks在上述第二步的所述的小波神经网络参数值范围内，选择n组小波神经网络参数，形成n个小波神经网络；利用上述第一步所述的训练样本库A对小波神经网络进行训练，再利用验证样本库B进行测试，得到不同小波神经网络的测试误差；Within the parameter value range of the wavelet neural network in the second step above, select n groups of wavelet neural network parameters to form n wavelet neural networks; use the training sample library A described in the first step to train the wavelet neural network , And then use the verification sample library B for testing to obtain the test errors of different wavelet neural networks;第四步：支持向量机网络训练，得到最优小波神经网络参数Step 4: Support vector machine network training to get the optimal wavelet neural network parameters基于上述第三步计算结果，得到了n组小波神经网络参数分别对应的测试误差；进而基于支持向量机网络训练可以得到最优的小波神经网络参数；Based on the calculation results of the third step above, the test errors corresponding to the n groups of wavelet neural network parameters are obtained; then the optimal wavelet neural network parameters can be obtained based on the support vector machine network training;第五步：基于小波神经网络和支持向量机的损伤监控Step 5: Damage monitoring based on wavelet neural network and support vector machine将地震作用下实测得到的位移最大值作为损伤后的代表值，并结合损伤前支座弹簧刚度与主梁位移计算值，代入上述已训练好的小波支持向量机网络，可以得到实际地震荷载作用下支座弹簧刚度的变化，从而对支座损伤状态进行识别。Taking the maximum displacement measured under the earthquake action as the representative value after the damage, combined with the calculated value of the spring stiffness of the support before the damage and the displacement of the main girder, and substituting the trained wavelet support vector machine network into the above-mentioned trained wavelet support vector machine network, the actual seismic load can be obtained The change of the spring stiffness of the lower support, so as to identify the damage state of the support.
- 根据权利要求1所述的基于小波神经网络和支持向量机桥梁地震损伤监控方法，其特征在于，第一步中基于D最优设计方法选取m组样本，并将 样本数据库分成训练样本库A和验证样本库B。The method for monitoring bridge seismic damage based on wavelet neural network and support vector machine according to claim 1, characterized in that, in the first step, m groups of samples are selected based on the D optimal design method, and the sample database is divided into training sample database A and Verify sample bank B.
- 根据权利要求1所述的基于小波神经网络和支持向量机桥梁地震损伤监控方法，其特征在于，第一步所述地震波时程数据可以随时计入最新的地震波记录，且训练样本库A为m-10组，验证样本库B为10组。The method for monitoring bridge seismic damage based on wavelet neural network and support vector machine according to claim 1, wherein the seismic wave time history data in the first step can be included in the latest seismic wave records at any time, and the training sample database A is m -10 groups, the verification sample bank B is 10 groups.
- 根据权利要求1所述的基于小波神经网络和支持向量机桥梁地震损伤监控方法，其特征在于，第二步支持向量机网络训练过程中，核函数选择RBF核函数，即k(x i,x)＝exp(-|x i-x| 2/2σ 2)。 The method for monitoring bridge seismic damage based on wavelet neural network and support vector machine according to claim 1, characterized in that, in the second step of support vector machine network training, the kernel function selects the RBF kernel function, namely k(x i ,x )=exp(-|x i -x| 2 /2σ 2 ).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| PCT/CN2019/106578 WO2021051332A1 (en) | 2019-09-19 | 2019-09-19 | Bridge seismic damage monitoring method based on wavelet neural network and support vector machine |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| PCT/CN2019/106578 WO2021051332A1 (en) | 2019-09-19 | 2019-09-19 | Bridge seismic damage monitoring method based on wavelet neural network and support vector machine |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| WO2021051332A1 true WO2021051332A1 (en) | 2021-03-25 |
Family
ID=74882980
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| PCT/CN2019/106578 WO2021051332A1 (en) | 2019-09-19 | 2019-09-19 | Bridge seismic damage monitoring method based on wavelet neural network and support vector machine |
Country Status (1)
| Country | Link |
|---|---|
| WO (1) | WO2021051332A1 (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115685323A (en) * | 2023-01-01 | 2023-02-03 | 北京云庐科技有限公司 | Bridge online real-time simulation computing system based on earthquake load monitoring |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104992248A (en) * | 2015-07-07 | 2015-10-21 | 中山大学 | Microgrid photovoltaic power station generating capacity combined forecasting method |
| CN105571638A (en) * | 2015-12-17 | 2016-05-11 | 安徽理工大学 | Machinery device fault combination prediction system and method |
| CN106874833A (en) * | 2016-12-26 | 2017-06-20 | 中国船舶重工集团公司第七0研究所 | A kind of mode identification method of vibration event |
| CN107860420A (en) * | 2016-09-22 | 2018-03-30 | 淮阴师范学院 | A kind of large and complex structure health monitoring systems based on multi-main body cooperation |
| WO2018165753A1 (en) * | 2017-03-14 | 2018-09-20 | University Of Manitoba | Structure defect detection using machine learning algorithms |
-
2019
- 2019-09-19 WO PCT/CN2019/106578 patent/WO2021051332A1/en active Application Filing
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104992248A (en) * | 2015-07-07 | 2015-10-21 | 中山大学 | Microgrid photovoltaic power station generating capacity combined forecasting method |
| CN105571638A (en) * | 2015-12-17 | 2016-05-11 | 安徽理工大学 | Machinery device fault combination prediction system and method |
| CN107860420A (en) * | 2016-09-22 | 2018-03-30 | 淮阴师范学院 | A kind of large and complex structure health monitoring systems based on multi-main body cooperation |
| CN106874833A (en) * | 2016-12-26 | 2017-06-20 | 中国船舶重工集团公司第七0研究所 | A kind of mode identification method of vibration event |
| WO2018165753A1 (en) * | 2017-03-14 | 2018-09-20 | University Of Manitoba | Structure defect detection using machine learning algorithms |
Non-Patent Citations (2)
| Title |
|---|
| YU FANHUA: "Research on Detecting Damage of Bridge Structure Based on Computational Intelligence", CHINESE DOCTORAL DISSERTATIONS FULL-TEXT DATABASE, 1 June 2008 (2008-06-01), pages 1 - 150, XP055792736, ISSN: 1674-022X * |
| ZHANG JI-YAN , BING ZHAO-HONG , ZHANG LI , LI FENG-MAN , TIAN LI: "Application Research of Artificial Intelligence in Fault Diagnosis", JOURNAL OF LIAONING UNIVERSITY(NATURAL SCIENCES EDITION), vol. 39, no. 3, 15 August 2012 (2012-08-15), pages 231 - 237, XP055792742, ISSN: 1000-5846, DOI: 10.16197/j.cnki.lnunse.2012.03.009 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115685323A (en) * | 2023-01-01 | 2023-02-03 | 北京云庐科技有限公司 | Bridge online real-time simulation computing system based on earthquake load monitoring |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108038300B (en) | Optical fiber state evaluation method based on improved membership function combined with neural network | |
| US11709979B1 (en) | Bridge damage identification method considering uncertainty | |
| CN101145214A (en) | Cable-stayed bridge cable damage positioning method based on modified reverse transmittance nerve network | |
| Xia et al. | Long‐term performance monitoring and assessment of concrete beam bridges using neutral axis indicator | |
| CN104200005A (en) | Bridge damage identification method based on neural network | |
| CN112610903A (en) | Water supply pipe network leakage positioning method based on deep neural network model | |
| CN112883497B (en) | Space valve reliability assessment method based on multi-source information fusion | |
| CN110417011A (en) | A kind of online dynamic secure estimation method based on mutual information Yu iteration random forest | |
| CN109558621B (en) | Structural damage identification method and system | |
| CN105930571A (en) | Unit temperature response monitoring value based correction method for finite element model of large-span steel bridge | |
| CN110210681A (en) | A kind of prediction technique of the monitoring station PM2.5 value based on distance | |
| CN111062170A (en) | Transformer top layer oil temperature prediction method | |
| CN112949131B (en) | Probability damage positioning vector method for continuous bridge cluster damage diagnosis | |
| CN110046409A (en) | A kind of components of steam turbine health state evaluation method based on ResNet | |
| CN112883522A (en) | Micro-grid dynamic equivalent modeling method based on GRU (generalized regression Unit) recurrent neural network | |
| CN114492164A (en) | Organic pollutant migration numerical model substitution method based on multi-core extreme learning machine | |
| WO2021051332A1 (en) | Bridge seismic damage monitoring method based on wavelet neural network and support vector machine | |
| CN111047476A (en) | Dam structure safety monitoring accurate prediction method and system based on RBF neural network | |
| CN105426665B (en) | Method is determined based on the DYNAMIC RELIABILITY of status monitoring | |
| Wang | Design of damage identification algorithm for mechanical structures based on convolutional neural network | |
| CN105741184A (en) | Transformer state evaluation method and apparatus | |
| CN117195647A (en) | Method, apparatus, device, medium and program product for post-earthquake evaluation of transformer bushings | |
| CN116842841A (en) | Immersed tunnel service state prediction method based on artificial intelligence | |
| Kowalski et al. | The comparison of linear models for PM10 and PM2. 5 forecasting | |
| CN114444798A (en) | Contact network operation and maintenance method and device based on space-time distribution technology |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| 121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 19946049 Country of ref document: EP Kind code of ref document: A1 |
|
| NENP | Non-entry into the national phase |
Ref country code: DE |
|
| 122 | Ep: pct application non-entry in european phase |
Ref document number: 19946049 Country of ref document: EP Kind code of ref document: A1 |