CN110631792A - Seismic hybrid test model updating method based on convolutional neural network - Google Patents

Seismic hybrid test model updating method based on convolutional neural network Download PDF

Info

Publication number
CN110631792A
CN110631792A CN201910965456.4A CN201910965456A CN110631792A CN 110631792 A CN110631792 A CN 110631792A CN 201910965456 A CN201910965456 A CN 201910965456A CN 110631792 A CN110631792 A CN 110631792A
Authority
CN
China
Prior art keywords
test
substructure
neural network
layer
convolutional neural
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN201910965456.4A
Other languages
Chinese (zh)
Other versions
CN110631792B (en
Inventor
王燕华
吴刚
王成
侯士通
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Southeast University
Original Assignee
Southeast University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Southeast University filed Critical Southeast University
Priority to CN201910965456.4A priority Critical patent/CN110631792B/en
Publication of CN110631792A publication Critical patent/CN110631792A/en
Application granted granted Critical
Publication of CN110631792B publication Critical patent/CN110631792B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
    • G01M7/00—Vibration-testing of structures; Shock-testing of structures
    • G01M7/02—Vibration-testing by means of a shake table
    • 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
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks

Landscapes

  • Physics & Mathematics (AREA)
  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Molecular Biology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (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)
  • Geophysics And Detection Of Objects (AREA)

Abstract

The invention discloses an earthquake-resistant hybrid test model updating method based on a convolutional neural network, which is characterized in that a system input variable and a restoring force observation value of a test substructure are used as a sample set to train the convolutional neural network on line, so that a numerical substructure restoring force prediction model which is more in line with the actual situation is obtained, and a restoring force model of the same or similar part of the numerical substructure as the test substructure is replaced. Therefore, selection errors of the model are avoided, the prediction precision of the restoring force is obviously improved, and the result of the hybrid test is more in line with the real situation. The method removes the pooling layer in the convolutional neural network, improves the calculation efficiency, and simultaneously keeps good data feature extraction capability and noise resistance capability. The prediction precision of the restoring force of the numerical substructure in the hybrid test is improved, the generalization capability and the noise resistance capability of the earthquake-resistant hybrid test model updating method based on the intelligent algorithm are obviously improved, and the modeling analysis result of the numerical substructure in the hybrid test is more accurate.

Description

Seismic hybrid test model updating method based on convolutional neural network
Technical Field
The invention relates to a structural earthquake-resistant performance evaluation test method in the field of civil engineering, in particular to an earthquake-resistant hybrid test model updating method based on a convolutional neural network.
Background
The general civil engineering field commonly used structural earthquake resistance test methods mainly have three types: pseudo-static tests, vibration table tests and pseudo-dynamic tests. The pseudo-static test is to carry out low-cycle repeated cyclic loading on a test piece according to certain load control or displacement control, so that the test piece is subjected to elastic stress till destruction, and a nonlinear constitutive model of a structure or a structural member is obtained. The pseudo-static test is most widely applied due to simple and stable technology, but has the defect that the influence of seismic waves on the structure cannot be considered. The earthquake simulation shaking table test can truly reproduce the earthquake action, but because the size and the bearing capacity of the earthquake simulation shaking table are limited, only a reduced-scale model test can be adopted. For the model test of a tall structure, all characteristics of the real structure under the action of earthquake cannot be reflected due to the small scale. The simulated dynamic test is an on-line test, the loading simulation is controlled by a computer to reproduce the seismic process, and the dynamic response loading restoring force and displacement are calculated according to a numerical integration algorithm. The method has the advantages that the restoring force characteristic of the structure in the simulated dynamic force test method is not derived from a mathematical model any more, but is directly measured from a test structure, the numerical error caused by assuming the restoring force model is avoided, the method can be applied to large-size model tests, and meanwhile, the gradual damage process of the structure can be observed in the test process.
The substructure mixing test was developed on the basis of the conventional pseudo-dynamic test method. For some large and complex structures, the substructure technology divides the structure into a test substructure and a numerical substructure, the part which is easy to damage or has complex nonlinear restoring force characteristics is used as the test substructure to carry out physical loading, the rest part is used as the numerical substructure to carry out numerical simulation in a computer, and the two parts are unified in the motion equation of the structure. The substructure technology has the advantages of facilitating the development of large engineering structure experiments and reducing the cost of test equipment and the scale of expenses.
When a large complex structure is subjected to a hybrid test. When the overall structure goes into non-linearity, it is not possible to perform physical loading tests on all critical parts, and some critical components or sites must be modeled in numerical sub-structures. However, the current substructure mixing test still has large model errors: on one hand, the model derived from the numerical substructure is too simplified to describe the nonlinear characteristics of the real structure; on the other hand, the uncertainty of the numerical substructure model parameters is caused, for example, if the assumed numerical model parameters are used to describe a member which cannot be tested in a large-scale complex structure and possibly enters nonlinearity, when the proportion of the numerical model is large, the precision of the whole mixed experiment is reduced, so that the test result cannot truly reflect the anti-seismic performance and the seismic response of the structure. In order to solve the problem of large error of a numerical substructure model in a substructure hybrid test, numerous scholars begin to research a model updating method of the numerical substructure in an earthquake-resistant hybrid test, and the model updating includes updating selected initial restoring force model parameters, not selecting an initial restoring force model and directly predicting the model. By using an intelligent algorithm, a restoring force model of the structure is not supposed in advance, and only sample data such as displacement, restoring force and the like of the test substructure are input into a network of the intelligent algorithm for training, so that the restoring force model which is more in line with the actual situation can be obtained. And the method can be further used for predicting the restoring force of the numerical substructure on line, updating the restoring force model obtained by training in real time according to the actual condition, and then carrying out the next mixed test loading process.
Disclosure of Invention
The purpose of the invention is as follows: the invention aims to provide an earthquake-resistant hybrid test model updating method based on a convolutional neural network, which can improve the accuracy of a numerical substructure resilience model in an earthquake-resistant hybrid test and improve the prediction capability and the anti-interference capability of the model on the resilience.
The technical scheme is as follows: in order to achieve the purpose, the invention adopts the following technical scheme:
the invention relates to an earthquake-resistant hybrid test model updating method based on a convolutional neural network, which comprises the following steps of:
a. dividing the overall structure into a test substructure and a numerical substructure, according to the number of degrees of freedom of the structure andstructural parameters, establishing a motion differential equation of the overall structure, and solving a target displacement d of the test substructure of the ith step of the hybrid testE，iThe obtained target displacement dE，iTransferring to a physical loading system, and pushing the test substructure to reach a target displacement d by an actuatorE，iObtaining the restoring force R of the test substructure by a sensor in the actuatorE，i；
b. Using the total system input variable { d } of the test sub-structure of j steps before the ith step and j +1 steps including the ithE，i-j，…，dE，iAnd observed values of restoring force of the test substructure { R }E，i-j，…，RE，iUsing the training sample set of the ith convolutional neural network as (d)E，i-j，RE，i-j)，…，(dE，i，RE，i) }; wherein d isE，iSystem input variable, R, representing the test substructure of step iE，iRepresenting the observed value of the restoring force of the test substructure in the ith step;
c. training sample { (d)E，i-j，RE，i-j)，…，(dE，i，RE，i) Firstly, processing the data through an input layer, entering a convolutional layer, obtaining the characteristics of the data through convolutional processing of a convolutional kernel, then entering an activation layer, wherein the activation layer comprises an excitation function for helping a training sample to express complex characteristics, finally, entering a full connection layer through the processed characteristics, and obtaining a prediction model through training a convolutional neural network
Figure BDA0002229448080000021
d. Using the prediction model obtained in step c
Figure BDA0002229448080000022
System input variable z of i-th step numerical substructure of mixed testiInputting a prediction model to obtain the restoring force R of the ith step numerical substructureN，iAnd R isN，iFeeding back to a numerical integration algorithm; thus, the mixing test of the step i is completed, and then the steps a-d are circulated until the seismic input is finished.
Furthermore, the loading device for loading the test substructure has an actuating signal from the integral solution of a differential equation of motion of each step of the structure, and the loading mode of the test substructure is a displacement control loading mode.
Further, in step b, the convolutional neural network comprises an input layer, a convolutional layer, an activation layer and a full link layer, and the pooling layer is removed.
Further, in the step c, the processing of the training samples by the input layer is normalized, that is, the data all become the range of 0 to 1; the training samples processed by the input layer are processed by convolution kernels in the convolution layer, the convolution kernels regularly sweep the input samples according to a certain step length, data points in a receptive field are multiplied point by point, obtained results are added and summed, finally, the sample characteristics of the training data are obtained, parameters of the convolution kernels are adjusted according to actual needs including the number and dimensionality of the input training samples, the obtained sample characteristics enter the activation layer, and the activation layer performs nonlinear mapping on the output results of the convolution layer; the finally processed training sample enters a full connection layer to carry out the training of a neural network; all neurons of the full connection layer are connected in a weight mode, and the full connection layer is located at the tail portion of the improved convolutional neural network.
Further, in step c, the activation functions in the activation layer are selected as two:
Figure BDA0002229448080000031
or
Figure BDA0002229448080000032
Further, in the step d, the restoring force R of the numerical substructureN，iIs a system input variable z of a numerical substructureiBy means of a prediction model
Figure BDA0002229448080000033
Processing the obtained yi(zi) As restoring force RN，i。
Furthermore, the neural network has no requirement on the size of the input training sample, can adopt various training sample sizes, can adjust the weight in the convolution kernel according to the requirement, and is suitable for different data types.
Has the advantages that: compared with the prior art, the invention has the following remarkable advantages:
(1) the convolutional neural network can well meet the requirement of updating a hybrid test model, the resilience of the numerical substructure at the current step is obtained by adopting the trained convolutional neural network model for prediction, the concrete situation is better met, and the noise of a training sample has stronger anti-interference capability; the method improves the prediction precision of the numerical value substructure resilience in the updating of the substructure hybrid test model, and obviously improves the precision of the hybrid test;
(2) the neural network can be well adapted to the change of the dimension of the input training sample, various training sample sizes can be adopted, and sudden change of the number of neurons in the neural network and sudden increase of the operation amount caused by the change of the size and the dimension of the input sample can be avoided; the weight in the convolution kernel can be adjusted according to the requirement, so that the requirements of different data types are met;
(3) the convolutional neural network provided by the invention has the advantages that the pooling layer is removed because the features extracted from the convolutional layer do not need to be compressed, the class calculation efficiency and speed are improved, and the convolutional neural network is more suitable for the condition of nonlinear fitting.
Drawings
FIG. 1 is a flow chart of a model update mixing test using the method according to an embodiment of the present invention;
FIG. 2 is a flowchart of an algorithm for a numerical substructure model updating method in accordance with an embodiment of the present invention;
FIG. 3 is a schematic diagram of a convolutional neural network in accordance with an embodiment of the present invention;
FIG. 4 is a schematic diagram of a convolutional neural network according to an embodiment of the present invention;
FIG. 5 is a diagram illustrating a convolution kernel processing training sample according to an embodiment of the present invention;
fig. 6 is a schematic diagram of a mixing test using the vibration table of the present invention.
Detailed Description
The technical solution of the present invention will be further described with reference to the following detailed description and accompanying drawings. Fig. 1 is a flow chart of a model updating hybrid test using the method in the embodiment of the present invention, and the embodiment discloses a method for updating a seismic hybrid test model based on a convolutional neural network, as shown in fig. 2, including the following steps:
a: and establishing a motion differential equation of the whole structure according to the number of the degrees of freedom of the structure and the structure parameters. Solving the target displacement d of the test substructure of the step i of the hybrid test by adopting a numerical integration algorithmE，iTarget displacement d of sum value substructureN，i. The obtained target displacement dE，iThe signal is converted into an electric signal and is transmitted to a controller to be converted into the actuating displacement of the hydraulic servo actuator, and the physical loading of the test substructure is realized. The actuator pushes the test substructure to reach the target displacement dE，iObtaining the restoring force R of the test substructure by a sensor in the actuatorE，i。
b: using the total system input variable { d } of the test sub-structure of j steps before the ith step and j +1 steps including the ithE，i-j，…，dE，iAnd observed values of restoring force of the test substructure { R }E，i-j，…，RE，iUsing the training sample set of the ith convolutional neural network as (d)E，i-j，RE，i-j)，…，(dE，i，RE，i) In which d isE，iThe system input variables representing the trial sub-structure of step i. RE，iAnd (4) representing the observed value of the restoring force of the test substructure in the ith step.
c: training sample { (d)E，i-j，RE，i-j)，…，(dE，i，RE，i) Firstly, processing the data by an input layer, then entering a convolution layer, and obtaining the characteristics of the data by convolution kernel convolution processing. And then into the activation layer, which contains a stimulus function to assist in expressing the complex features. Finally, the processed features enter a full connection layer, and a prediction model is obtained through training
Figure BDA0002229448080000041
d: using the prediction model obtained in step c
Figure BDA0002229448080000042
System input variable z of i-th step numerical substructure of mixed testiInputting a prediction model to obtain the restoring force R of the ith step numerical substructureN，iAnd R isN，iAnd feeding back to a numerical integration algorithm. Thus, the mixing test of the step i is completed, and then the steps a-d are circulated until the seismic input is finished.
In the step a, the actuating signal of the loading equipment for loading the test substructure comes from the integral solution of a motion differential equation of each step of the structure, and the loading mode of the test substructure is a displacement control loading mode. In the step b, the convolutional neural network is based on the standard convolutional neural network, and the features extracted from the convolutional layer do not need to be compressed, so that the pooling layer is removed, the calculation efficiency is improved, and the convolutional neural network is more suitable for the condition of nonlinear fitting.
The schematic diagram of the convolutional neural network involved in step c is shown in fig. 3, and the structural schematic diagram of the convolutional neural network is shown in fig. 4, and the convolutional neural network mainly comprises an input layer, a convolutional layer, an active layer and a full connection layer. And (4) forming training samples by data obtained by the experiment, inputting the training samples into the convolutional neural network, and forming a matrix. The training samples were entered into the convolutional layer. The convolution kernel will sweep through the training sample according to a certain rule to form a receptive field. Then, convolution processing is performed on the data points in the receptive field, as shown in fig. 5, where a ', b ', c ', d ', e ', f ', g ', h ', i ' are weights in a convolution kernel, and α, β, a ', β ' are training samples output after the convolution kernel processing.
α＝1×a+2×b+3×c+4×δ+5×e+6×f+7×g+8×h+9×i
β＝4×a+5×b+6×c+7×d+8×e+9×f+10×g+11×h+12×i
The size of the convolution kernel can be determined according to actual conditions. When the convolution process of the current step is finished, the convolution sum slides forwards by an artificially defined step size and carries out the next convolution processThe process. A layer of training samples is obtained that reveal data features that are integrated as shown in fig. 5. RE，i-1·ΔdE，iFor the energy consumption of the structure of step i, Ei-1Represents the cumulative energy consumption of the structure of step i-1, Ei-1＝Ei-2+|dE，i-1·RE，i-1L. And then into the active layer. Considering the fitting capability to introduce the non-linear function, the activation functions chosen in the activation layer are two of:
(1) RELU function:
Figure BDA0002229448080000051
the convergence rate is high, and the gradient is simple to calculate;
(2) tan h function:
Figure BDA0002229448080000052
during the calculation, the RELU function is preferably selected, and the tanh function is selected if the effect is not ideal.
The activation layer can help the sample to better display the nonlinear characteristics through the mapping of the activation function. And then, the training data finally reach a full connection layer, and all neurons between each layer of the full connection layer are connected with weights and are positioned at the tail part of the convolutional neural network. Training the convolutional neural network to finally obtain a restoring force model which is more in line with the actual situation
Figure BDA0002229448080000053
The method is described in detail below with reference to specific examples, and as shown in fig. 6, the method is a shaking table hybrid test using the convolutional neural network model updating method disclosed in the present invention. In the vibration table mixing test, a large error exists in the modeling analysis of the numerical substructure due to model errors. The test data of the test sub-structure may be used to update a computational model of similar structures or components in the numerical sub-structure. However, the data generated in the hybrid test process of the vibration table contains more noise, and the requirements of model prediction and noise resistance are met at the same time, so that the method based on the noise prediction and the noise resistance provided by the invention can be adoptedA mixed test model updating method of a convolutional neural network. M in FIG. 6i，Ki，Ci(i ═ 1, 2, 3) are the structural parameters mass, stiffness and damping respectively. MN，CNRespectively the mass and damping matrix of the numerical substructure. a iskIs the acceleration, vkIs the velocity.
Figure BDA0002229448080000061
To derive the restoring force from the existing restoring force model,
Figure BDA0002229448080000062
is the restorative force predicted by the convolutional neural network-based model update module. RE，kIn order to test the restoring force of the substructure,
Figure BDA0002229448080000063
displacement command of the test substructure input to the controller, obtained for numerical integration calculation, FkFor external excitation, here seismic action, typically
Figure BDA0002229448080000064
Where m is the matrix of the structural mass,for seismic acceleration, k, (k ═ 1, 2, …) is the number of steps integrated for each step value in the hybrid test.
The specific implementation steps are as follows:
1. and extracting the part with the inaccurate analysis result which is difficult to model and analyze in the whole structure as a test substructure. The rest is taken as a numerical substructure. Carrying out modeling analysis on the numerical substructure, and carrying out a real physical loading test on the test substructure;
2. establishing a motion differential equation of the whole structure, wherein the speed, the displacement and the acceleration of the structure are zero at the beginning of the test, and the displacement and the acceleration of the vibration table are also equal to 0;
3. assuming that k is 1, the restoring force R of the substructure is testedE，1Number is calculated when it is 0Value substructure of seismic inertia force and RE，1Under combined action of displacement, velocity and accelerationAnd will signalTransmitting the driving command to the actuator and the vibration table, and measuring the restoring force of the test substructure at the end of the first step
Figure BDA0002229448080000068
And transmits the data to a computer and feeds back the data to a numerical integration algorithm;
4. calculating the seismic action and the experimental substructure restoring force in the k-th step (k-2, 3, 4 … …)
Figure BDA0002229448080000069
Under the combined action of the two, the structure at the junction of the numerical substructure and the test substructure dynamically responds to displacement, speed and acceleration
Figure BDA00022294480800000610
Simultaneously acquiring displacement u of vibrating tablekAnd will be
Figure BDA00022294480800000611
As a drive command to the actuator.
5. Actuator application
Figure BDA00022294480800000612
Simultaneously inputting the seismic acceleration of the ith step to a vibration table in the test substructure, and measuring by a sensor in an actuator to obtain the actual displacement of the test substructure at the end of the kth step
Figure BDA00022294480800000613
And reaction force of the test substructure
Figure BDA00022294480800000614
Feeding back to computer to formTraining sample set
Figure BDA00022294480800000615
6. Training sample
Figure BDA0002229448080000071
Firstly, the input layer is processed, then the input layer enters the convolution layer, and the convolution kernel convolution processing is used for obtaining the characteristics of data. And then entering an activation layer, wherein the activation layer comprises an excitation function to express complex characteristics, finally entering a full connection layer through the processed characteristics, training the convolutional neural network, and finally obtaining a prediction model
7. Using the obtained prediction model
Figure BDA0002229448080000073
Inputting the k step system input variable of the part similar or identical to the test substructure in the numerical substructure
Figure BDA0002229448080000074
Inputting the prediction model to obtain the predicted value of the restoring force of the part in the kth step numerical value substructure
Figure BDA0002229448080000075
Will obtain
Figure BDA0002229448080000076
And feeding back to a numerical integration algorithm. The rest part of the numerical substructure adopts the existing restoring force model, and the displacement obtained by calculating the numerical substructure
Figure BDA0002229448080000077
Inputting restoring force model to obtain restoring force
Figure BDA0002229448080000078
And performing numerical substructure calculation. Thus, the mixing test of the k step is completed, and then the steps 1 to 7 are circulated until the earthquake occursAnd after the dynamic input is finished, the dynamic response of the whole structure can be solved finally.

Claims (7)

1. A method for updating an earthquake-resistant hybrid test model based on a convolutional neural network is characterized by comprising the following steps: the method comprises the following steps:
a. dividing the integral structure into a test substructure and a numerical substructure, establishing a motion differential equation of the integral structure according to the number of degrees of freedom and structural parameters of the structure, and solving a target displacement d of the test substructure at the ith step of the hybrid testE，iThe obtained target displacement dE，iTransferring to a physical loading system, and pushing the test substructure to reach a target displacement d by an actuatorE，iObtaining the restoring force R of the test substructure by a sensor in the actuatorE，i；
b. Using the total system input variable { d } of the test sub-structure of j steps before the ith step and j +1 steps including the ithE，i-j，…，dE，i) And observed values of restoring force { R } of the test substructuresE，i-j，…，RE，iUsing the training sample set of the ith convolutional neural network as (d)E，i-j，RE，i-j)，…，(dE，i，RE，i) }; wherein d isE，iSystem input variable, R, representing the test substructure of step iE，iRepresenting the observed value of the restoring force of the test substructure in the ith step;
c. training sample { (d)E，i-j，RE，i-j)，…，(dE，i，RE，i) Firstly, processing the data through an input layer, entering a convolutional layer, obtaining the characteristics of the data through convolutional processing of a convolutional kernel, then entering an activation layer, wherein the activation layer comprises an excitation function for helping a training sample to express complex characteristics, finally, entering a full connection layer through the processed characteristics, and obtaining a prediction model through training a convolutional neural network
d. Using the prediction model obtained in step c
Figure FDA0002229448070000012
System input variable z of i-th step numerical substructure of mixed testiInputting a prediction model to obtain the restoring force R of the ith step numerical substructureN，iAnd R isN，iFeeding back to a numerical integration algorithm; thus, the mixing test of the step i is completed, and then the steps a-d are circulated until the seismic input is finished.
2. The convolutional neural network-based seismic hybrid test model updating method as claimed in claim 1, wherein: the loading device for loading the test substructure has an actuating signal from integral solution of a motion differential equation of each step of the structure, and the loading mode of the test substructure is a displacement control loading mode.
3. The convolutional neural network-based seismic hybrid test model updating method as claimed in claim 1, wherein: in step b, the convolutional neural network comprises an input layer, a convolutional layer, an activation layer and a full-link layer, and the pooling layer is removed.
4. The convolutional neural network-based seismic hybrid test model updating method as claimed in claim 1, wherein: in the step c, the input layer normalizes the training samples, namely, the data are all changed into the range from 0 to 1; the training samples processed by the input layer are processed by convolution kernels in the convolution layer, the convolution kernels regularly sweep the input samples according to a certain step length, data points in a receptive field are multiplied point by point, obtained results are added and summed, finally, the sample characteristics of the training data are obtained, parameters of the convolution kernels are adjusted according to actual needs including the number and dimensionality of the input training samples, the obtained sample characteristics enter the activation layer, and the activation layer performs nonlinear mapping on the output results of the convolution layer; the finally processed training sample enters a full connection layer to carry out the training of a neural network; all neurons of the full connection layer are connected in a weight mode, and the full connection layer is located at the tail portion of the improved convolutional neural network.
5. The convolutional neural network-based seismic hybrid test model updating method as claimed in claim 4, wherein: in step c, the activation functions in the activation layer are selected as two:
Figure FDA0002229448070000021
or
Figure FDA0002229448070000022
6. The convolutional neural network-based seismic hybrid test model updating method as claimed in claim 1, wherein: in said step d, the restoring force R of the numerical substructureN，iIs a system input variable z of a numerical substructureiBy means of a prediction model
Figure FDA0002229448070000023
Processing the obtained yi(zi) As restoring force RN，i。
7. The convolutional neural network-based seismic hybrid test model updating method as claimed in claim 1, wherein: the neural network has no requirement on the size of an input training sample, can adopt various training sample sizes, can adjust the weight in the convolution kernel according to the requirement, and is suitable for different data types.
CN201910965456.4A 2019-10-11 2019-10-11 Seismic hybrid test model updating method based on convolutional neural network Active CN110631792B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910965456.4A CN110631792B (en) 2019-10-11 2019-10-11 Seismic hybrid test model updating method based on convolutional neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910965456.4A CN110631792B (en) 2019-10-11 2019-10-11 Seismic hybrid test model updating method based on convolutional neural network

Publications (2)

Publication Number Publication Date
CN110631792A true CN110631792A (en) 2019-12-31
CN110631792B CN110631792B (en) 2021-01-05

Family

ID=68975978

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910965456.4A Active CN110631792B (en) 2019-10-11 2019-10-11 Seismic hybrid test model updating method based on convolutional neural network

Country Status (1)

Country Link
CN (1) CN110631792B (en)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111257934A (en) * 2020-01-17 2020-06-09 哈尔滨工业大学 Seismic oscillation peak acceleration prediction method based on second-order neuron deep neural network
CN112362276A (en) * 2020-10-27 2021-02-12 南京林业大学 Substructure mixing test method
CN112380631A (en) * 2020-12-02 2021-02-19 黑龙江科技大学 Novel iterative hybrid test method based on neural network
CN112861383A (en) * 2021-03-17 2021-05-28 哈尔滨工业大学 Railway station anti-seismic toughness evaluation method and system
CN115796038A (en) * 2022-12-02 2023-03-14 哈尔滨工业大学 Real-time hybrid test method based on recurrent neural network

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2006201089A (en) * 2005-01-21 2006-08-03 Toyota Motor Corp Model characteristic generation method
KR20150035633A (en) * 2013-09-27 2015-04-07 한국전력공사 Apparatus for measuring earthquake intensity and method for the same
CN108460152A (en) * 2018-03-26 2018-08-28 王智华 A kind of the space-filling curve method, apparatus and computer readable storage medium of data
CN108520277A (en) * 2018-04-09 2018-09-11 哈尔滨工业大学 Reinforced concrete structure seismic Damage automatic identification based on computer vision and intelligent locating method
CN109031415A (en) * 2018-06-20 2018-12-18 清华大学 A kind of controlled source data ring drawing method based on depth convolutional neural networks
CN109885916A (en) * 2019-02-02 2019-06-14 东南大学 A kind of bulk testing on-time model update method based on LSSVM
CN110032975A (en) * 2019-04-15 2019-07-19 禁核试北京国家数据中心 A kind of pick-up method of seismic phase
CN111580151A (en) * 2020-05-13 2020-08-25 浙江大学 SSNet model-based earthquake event time-of-arrival identification method

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2006201089A (en) * 2005-01-21 2006-08-03 Toyota Motor Corp Model characteristic generation method
KR20150035633A (en) * 2013-09-27 2015-04-07 한국전력공사 Apparatus for measuring earthquake intensity and method for the same
CN108460152A (en) * 2018-03-26 2018-08-28 王智华 A kind of the space-filling curve method, apparatus and computer readable storage medium of data
CN108520277A (en) * 2018-04-09 2018-09-11 哈尔滨工业大学 Reinforced concrete structure seismic Damage automatic identification based on computer vision and intelligent locating method
CN109031415A (en) * 2018-06-20 2018-12-18 清华大学 A kind of controlled source data ring drawing method based on depth convolutional neural networks
CN109885916A (en) * 2019-02-02 2019-06-14 东南大学 A kind of bulk testing on-time model update method based on LSSVM
CN110032975A (en) * 2019-04-15 2019-07-19 禁核试北京国家数据中心 A kind of pick-up method of seismic phase
CN111580151A (en) * 2020-05-13 2020-08-25 浙江大学 SSNet model-based earthquake event time-of-arrival identification method

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111257934A (en) * 2020-01-17 2020-06-09 哈尔滨工业大学 Seismic oscillation peak acceleration prediction method based on second-order neuron deep neural network
CN111257934B (en) * 2020-01-17 2022-03-11 哈尔滨工业大学 Seismic oscillation peak acceleration prediction method based on second-order neuron deep neural network
CN112362276A (en) * 2020-10-27 2021-02-12 南京林业大学 Substructure mixing test method
CN112362276B (en) * 2020-10-27 2022-04-15 南京林业大学 Substructure mixing test method
CN112380631A (en) * 2020-12-02 2021-02-19 黑龙江科技大学 Novel iterative hybrid test method based on neural network
CN112380631B (en) * 2020-12-02 2023-02-14 黑龙江科技大学 Novel iterative hybrid test method based on neural network
CN112861383A (en) * 2021-03-17 2021-05-28 哈尔滨工业大学 Railway station anti-seismic toughness evaluation method and system
CN112861383B (en) * 2021-03-17 2022-09-16 哈尔滨工业大学 Railway station anti-seismic toughness evaluation method and system
CN115796038A (en) * 2022-12-02 2023-03-14 哈尔滨工业大学 Real-time hybrid test method based on recurrent neural network

Also Published As

Publication number Publication date
CN110631792B (en) 2021-01-05

Similar Documents

Publication Publication Date Title
CN110631792B (en) Seismic hybrid test model updating method based on convolutional neural network
US8095344B2 (en) Methods and systems for modeling material behavior
CN109885916B (en) Mixed test online model updating method based on LSSVM
CN110795884B (en) Novel hybrid test method based on multi-scale model updating
CN115577436B (en) Combined deep learning method for solving wind-induced vibration response of uncertain structure
CN111368466B (en) Mechanical vibration prediction method based on frequency response function parameter correction
CN109902404A (en) The unified recurrence calculation method of the structure time-histories data integral of different damping form
CN110000787A (en) A kind of control method of super redundant mechanical arm
JP3882014B2 (en) Structure vibration test apparatus and vibration test method therefor
CN106017953B (en) A kind of test method and system for large and complex structure experimental study
CN111027261B (en) Hybrid simulation test method for researching structural wind excitation response
Yang et al. Hybrid simulation of a zipper‐braced steel frame under earthquake excitation
Schumacher et al. Simulation-ready characterization of soft robotic materials
Duan et al. A technique for inversely identifying joint stiffnesses of robot arms via two-way TubeNets
CN112362276B (en) Substructure mixing test method
CN110765560B (en) Mechanical mechanism vibration prediction method based on time-varying damping
CN106844991B (en) Air spring rigidity self-balancing iterative identification method for air floating type vibration control system
CN115796038B (en) Real-time hybrid test method based on cyclic neural network
CN104008234B (en) Method for correcting closely spaced mode model with damping structure
CN115794644A (en) Real-time hybrid test method based on single-test-piece restart multi-task loading
Mueller Real-time hybrid simulation with online model updating
CN112487689A (en) Mixed test method based on statistical CKF model updating
Hosseini et al. A framework for multi‐element hybrid simulation of steel braced frames using model updating
Ghaffary et al. A hybrid simulation approach for aeroelastic wind tunnel testing: challenges and foundational work
CN116738802A (en) Real-time hybrid test method and system based on parameter identification and deep learning agent model

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant