Research And Application Of Damage Identification Method Based On Improved Uniform Design Table

In the long-term use,the structure will inevitably produce some damages due to environmental factors such as corrosion,wind,earthquake and human-made damage.If the damages can not be repaired in time,it will be accumulated in the use process,and the damage will gradually increase,which will reduce the safety and reliability of the structure.The damage of essential components will even lead to sudden damage of the structure.Accurately identify the structural damage in the early stage of structural damage is of considerable significance to ensure the safety and reliability of the structure and provide the repair scheme of the structure.In the present research results,by using the uniform design table to carry out the simulation experiment,we can build the correlation database between the different damage conditions of the structure and the change of the identification index.After mapping the damage database with neural network tools,it can be used to identify the damage location and degree of the structure.However,the uniform design test method takes too much account of the uniformity of the test,so it cannot carry out multiple tests for essential factors.This method reduces the number of test samples,and it cannot construct a complete and useful database of neural network training samples.Besides,in the process of recognition,the neural network has some problems,such as training non convergence and training error tends to local minimum,and the recognition index is not sensitive to the damage location and degree.Because of the above problems,the main contents of this paper are as follows.First of all,in order to determine the complex mapping between structural damage and structural response,this paper uses the characteristic of artificial neural network,such as self-learning,self-adaptive,associative memory et al.Based on the Matlab toolbox,the Generalized Regression Neural Network(GRNN)and Radial Basis Function(RBF)Neural Network suitable for pattern classification and nonlinear fitting are established respectively.On this basis,a two-step damage identification method is proposed.Firstly,the damage location of the structure is determined by the GRNN,and then the damage degree of the structure is identified by RBF Neural Network.The validity of the two-step recognition method is verified by the numerical simulation analysis of the frame bent structure of large thermal power plant.Secondly,in order to find a useful damage index which is sensitive to both the damage location and damage degree of the structure,the change rate of the modal strain energy of the plane element is proposed as the damage identification index of the structure through the research and formula derivation of the energy change before and after the damage of the structure.This index overcomes the defect that the spatial structure cannot obtain the complete modal signal of the structure,Through the numerical simulation of a nine floors spatial frame structure,the effectiveness of the damage index to identify the damage location and damage degree of the structure is verified.Thirdly,in order to reduce the number of neural network forward analysis corrections and consider the relative importance of different components of the structure.In this paper,the orthogonal design method is used to optimize the uniform design table,and then the analytic hierarchy process is used to evaluate the importance of structural components.The uniform design experiment is optimized by different density selection,and the sample database for training generalized regression neural network is selected with emphasis and representative.Through the above method,the damage location of the structure can be accurately identified with a large number of reduced training samples,which has a higher recognition efficiency.Through the numerical simulation of a large-scale frame structure,it is verified that it can still have a good recognition accuracy and reduce the operation cost in the case of reducing a large number of neural network training samples.Finally,through the numerical simulation and experimental analysis of a four floors steel frame structure under different working conditions of single damage and multiple damage,the effectiveness of the proposed damage identification method is verified.On this basis,a complete identification process for the damage identification method is proposed,and the effectiveness of the identification process is verified by the application of damage identification to a steel truss bridge and a nine story frame structure,This is the basis for the application of this method to the damage identification of different engineering structures.