Font Size: a A A

The Multi-hierarchical Damage Detection Method Of The Frame Structure Based On Uniform Design Experiment

Posted on:2019-04-02Degree:MasterType:Thesis
Country:ChinaCandidate:W Q TangFull Text:PDF
GTID:2322330542463822Subject:Architecture and civil engineering
Abstract/Summary:PDF Full Text Request
The frame structure has been widely used in design of domestic and foreign architectures,because of its advantages of economical cost,flexible space layout and good durability.Nevertheless,it is inevitable that structural damage could appear among the long-term use,and the unrepaired damage will accumulate and develop into more serious damage,which even causes sudden structural destruction.Thus,the survey of how to identify damage of structure is of great significance to prolong the life of structure and maintain the safety of people's life and property.On the basis of previous researches,the neural network with uniform design experiment method is used to establish the experiment table,which can be used for the forward analysis about damage identification,and the results show this approach used in this article is theoretically feasible.First of all,there are some features of BP neural network,such as abilities of learning,induction,summary,association and memory and strong robustness,which could be used to approximate the complex mapping between damage state and structural responses in the process of damage identification.During the process,the BP neural network is established by the MATLAB neural network toolbox,and L-M optimization algorithm is applied.Training function trainlm is used to realize the adaptive learning,and the ASCE SHM benchmark model has carried on the damage identification whose results prove the effectiveness of the constructed neural network.Secondly,in order to find effective index for damage identification of frame structure,the effectiveness of the proposed damage identification parameters has been proved by using normalized frequency change rate and displacement relative change rate,through a derivation of different damage identification parameters,and an example of frame structure.By using multi-hierarchical identification method of column frame structure damage location and degree,the damage identification of the components can be divided into three steps.It is the first step that structural normalized frequency change rate is used to locate the damage floors.The second step is to use the displacement relative change rate to locate the damage members.Third step,combining with damage degree results obtained by two identification parameters,the effectiveness of the multi-hierarchical damage identification has been illustrated by Numerical simulation of a concrete frame structure.Then,for reducing the number of times of positive analysis model correction and optimizing neural network training samples,uniform design method is used to establish the experiment table,which can be used for the forward analysis about damage identification.BP neural network is trained by the damage locations and extent that sampled from the experiment table.By comparing with the complete sample training examples,the method of uniform design for modifying the model could guarantee the training accuracy and reduce the number of test times.Finally,aim to verify the applicability of different frame structures based on uniform design experiments,the validity of this method is corroborated by irregular frame structure,frames with variable cross sections and examples of structural damage identification of finite element model of steel frame structure.In order to avoid deviation of large-scale uniform design table,the combination of partition step identification and small uniform design has been applied in the irregular framework example,which could divide the structure into multiple regions to construct training samples by uniform design forms.The result shows the uniform design experiment could use less training times of the neural networks to obtain the similar detecting performance;the multi-hierarchical identification method based on uniform design experiment can accurately capture the damage member of the frame structures,which could avoid defects of the single step or single parameter identification method.Based on the above analysis and calculation example,the method of the frame structure damage identification based on uniform design experiment has been proved to be feasible and effective.And the relevant analysis lays the theoretical foundation for later test verification.
Keywords/Search Tags:damage detection, BP neural network, frame structure, uniform design experiment, forward analysis
PDF Full Text Request
Related items