A New Nonlinear System Identification Method With Time Domain Measurements

There are a large number of non-linear components in civil engineering structures.On the one hand,under the action of earthquake,explosion,impact and other loads,the whole structure or components often show strong non-linearity.On the other hand,with the improvement of science and technology and people's living standards,people put forward higher requirements for disaster prevention and mitigation ability of civil engineering structures.More and more non-linear components such as dampers and buckling braces are applied to civil engineering structures.The working state of these components or devices directly affects the safety of civil engineering structures under extreme disasters.Therefore,based on the monitoring data of dynamic response of engineering structures under strong dynamic loads,the working status of these non-linear components and devices can be quantitatively evaluated by system identification,which is of great significance for understanding the health status of structures and rapid evaluation of building components after disasters.However,the traditional time-domain damage identification method has the limitation that the measurement noise seriously affects the identification accuracy.In order to overcome this limitation,this paper begins with the identification of modal damping ratio of linear elastic structures and studies the identification method of linear or non-linear systems based on incomplete time-domain signals,taking into account the correlation between different measured signals.By means of theoretical analysis and derivation,numerical simulation and experimental verification,the following aspects are thoroughly studied in this paper:(1)For linear elastic structures,this paper improves the method of sensitivity analysis to identify modal damping ratios based on time domain signal principal component analysis.At first the principal component analysis is used to determine the high quality principal component and projection space for some measured acceleration responses.Then project time domain signal sensitivity parameter identification equation to subspace corresponding to high quality principal component,identifying damping ratios with the projection equation to improve the anti-noise ability.The effectiveness and practicability of the method are verified by numerical simulation of plane truss structure and test of plane steel frame structure.(2)For structures with non-linear components,in order to identify the parameters of MR dampers in structures,Newmark method and Newton-Raphson iteration are used to establish a parameter identification method based on the sensitivity of time-domain response signals,which preliminarily solves the problem that the sensitivity parameter identification method has been confined to simple non-linear systems of multi-degree-offreedom for a long time.Then,principal component analysis and subspace projection,which may be helpful to fusing multi-sensor signals to identify structural parameters,are used to improve the anti-noise ability and stability of parameter identification.Based on the recognition results,it is proposed that the noise influence coefficient of the principal component is the most important index for the selection of high quality components.The accuracy of the method is verified by numerical simulation of multi-degree-of-freedom system and plane frame structure.(3)For structures with non-linear components,the explicit Newmark method is proposed to identify the non-linear restoring force imposed on the structure by partial measurement responses,without considering the complex non-linear behavior inside the components.The position effect and anti-noise ability of the method are compared with the explicit Runge-Kutta force identification method.At last the non-linear restoring force applied to the target substructure is also identified by using the interface force of the substructure and part of the measured responses.The accuracy and validity of the above methods are verified by numerical simulation of five-degree-of-freedom system and tendegree-of-freedom system respectively.