Structural Damage Identification Based On Time Series Model Under Environmental Excitation

In recent years,with the rapid development of sensor technology,structural health monitoring(SHM)is becoming an important part of the operation and management of basic civil facilities.Structural inspection engineers perform maintenance repairs by assessing structural damage to reduce unnecessary personnel and property damage.At present,global detection technology based on structural vibration characteristics has become the core of the SHM field.In the global detection technology based on structural vibration characteristics,the method of using environmental excitation to discriminate structural damage has become a hot topic in this research.The non-modal parameter method directly uses the data of structural vibration response or through some transformation from the measured response data.Extract information that can reflect structural damage.Compared with the structural modal parameter method,the non-modal parameter method is convenient,fast and accurate,and can better meet the real-time requirements of structural health monitoring.Time series analysis is a feature extraction method based on time domain analysis,which has recently received more and more attention from researchers in the field of SHM.However,when applying the time series model,this paper encounters the problem that the AR coefficient of the time series model is unstable.For the approximation signal with only minor differences and the time series model analysis,the extracted AR coefficients are very unstable.After derivation,it was found that the problem of the ill-conditioned matrix of the least squares equation when solving the coefficients was caused.Finally,the problem of this ill-conditioned matrix is eliminated by adding a regularization coefficient.Therefore,a regularization analysis of the time series model is proposed.This time series model with regularization analysis is called a regularized time series model.In order to solve the problem of structural damage identification under environmental excitation,this paper uses the regularized time series model to analyze the non-modal parameter method of a three-degree-of-freedom chain structure model.The structural vibration response is extracted under the fixed mode excitation and white noise mode excitation,and the structural vibration response is directly analyzed by the regularized time series model.The experimental results show that the regularized time series model can identify structural damage in the case of fixed mode excitation.However,in the case of white noise mode excitation,the damage index is unstable.In the damage detection of large structures,it is most convenient to use environmental excitation.Obviously,the regularized time series model can not directly identify the damage caused by the structure under environmental excitation.The virtual impulse response function is extended in the environmental load excitation technology.It can not only effectively overcome the randomness and instability of environmental excitation,but also represent the inherent characteristics of the structure.In order to verify this characteristic of the virtual impulse response function,it is combined with the signal feature extraction method of wavelet packet energy entropy to perform damage identification analysis on the same three-degree-of-freedom chain structure model.Ten random white noise excitations were performed on the three-degree-of-freedom chain structure.The effectiveness of the method is verified by comparing the results of 10 random excitations.The experimental results show that the combination of the virtual impulse response function and the wavelet packet energy entropy can identify the damage of the structure under environmental excitation,but the recognition effect on minor damage is not good.Moreover,the response amplitude of the three-degree-of-freedom chain structure is large,which does not conform to the dynamic amplitude of the response collected by the actual structure.The wavelet packet analysis method is suitable for random signals.The time domain feature extraction method for time series model analysis is more suitable for feature extraction of non-random signals,and the virtual impulse response function is in the chain structure under certain conditions.Random signals,so time series model analysis is more suitable for the study of the virtual impulse response function under certain conditions.This paper combines the virtual impulse response function with regularized time series model analysis for the first time.Applying this new method to the three-degree-of-freedom chain structure mentioned above,it is also subjected to 10 random excitations.The results of the run can be found that the method of time series model analysis is more sensitive to minor damage.In order to verify that the method is also effective when the amplitude of the acquisition structure signal is small,the method is applied to a four-degree-of-freeness plate column structure with a small response amplitude.The experimental results show that the method is still effective when the amplitude of the acquired signal is small.In order to verify the practicability of the method,this new method was finally applied to Alamos' four-layer frame structure experiment.The recognition results also verified the effectiveness of the method and its robustness to environmental excitation.