Vibration Feature Extraction And Structure Damage Detection In Heavy Noise

In the process of damage detection of engineering structures based on vibration response,the vibration response signal not only contains the structural damage information,but also has a large amount of noise.The presence of noise will seriously affect the data extraction effect of structural vibration response,especially for the initial small damage state when the structural response characteristics are relatively weak.It is difficult to obtain good results without the proper method of noise reduction.Therefore,this paper mainly studies on vibration response based signal feature extraction and damage location under strong background noise.The research of vibration feature extraction and damage detection are studied,and the work that is done can be showas follows:Large ParameteradaptiveStochastic Resonance that based on twice sampling is proposed for the purpose of eliminating the interference of noise in rotating machinery rubbing rotor fault detection.This Algorithm can expand to the application of traditional stochastic resonance algorithm which is only suitable for small-parameter condition.Simulation and experimental data analysis results show that the method can significantly improve the signal to noise ratio of the test data.This dissertation applies Duffing chaotic oscillator into structure damage detection and a new damage detection method based on Duffing chaotic oscillator and response sensitivity is proposed.A set of specific parameters are selected for the extraction of response signal,this avoids the cumbersome process of the traditional Duffing oscillator parameter's selection.The method is applied in the threedimensions beam model and bridge model,the results show that this method can well realize structural damage location under strong background noise,and this can provide a better methodology for structural damage identification.In order to achieve the precise identification of test parameters,the combination of phase space reconstruction and singular spectrum decomposition is conducted when supersonic flight test data under aerodynamic noise is filtered.Firstly,the feasibility of the method is demonstrated by numerical simulation.Then,the phase space reconstruction of the test data is reconstructed by the acoustic vibration test of a certain type of supersonic UAV,and the singular value decomposition is performed to obtain the signal subspace and the noise subspace which reflects the real signal information.By defining the singular value difference spectrum to determine the true signal subspace dimension,and for the existing maximum difference spectrum theory,we propose a method to optimize the differential spectrum peak value.The reconstruction results show that the method is suitable for the data processing of the acoustic vibration test of the aircraft under the supersonic flight condition.Consider the noise interference in sheet structure(aluminum plate)damage detection process,a method of damage detection that based on singular spectrum analysis and maximum likelihood principle is proposed.Singular spectrum analysis is firstly applied to the Lamb wave response signals,andoptimizing difference spectrum theory is adopted for reconstruction.Genetic Algorithm that based on the principle of maximum likelihood is applied for the optimization.The aluminum test results show that the method.is practical and effective.This dissertation also discusses nonlinear structure damage detection under strong background noise,and the corresponding physical equivalent model is designed for the nonlinear mass-spring system.Physical equivalence is mainly implemented by changing the original nonlinear system into an enhanced linear system,and the degree of freedom is increased for this enhanced linear system.The dynamic equilibrium equations of nonlinear term are built.Singular spectrum analysis,direct parameter identification method and general matrix rank perturbation theory are then adopted for damage location and damage extent determination and good identification results are obtained.