Pattern Recognition And Parameter Identification Of Nonlinear Element Based On Structural Dynamic Responses

Civil structures often perform strong nonlinear dynamic behaviors when subjected to extreme loads,i.e.earthquakes,typhoons,rain and snow loads et al.The nonlinear behaviors can be the large deformation,the yield of materials,the changing of boundary conditions.For a nonlinear structure,the traditional linear model calibration and validation approaches may not be applicable.Therefore,this paper proposes an approach for the pattern recognition and parameter identification of nonlinear structures based on the structural dynamic response.In this thesis,the dynamic characteristics and restoring force surfaces of a nonlinear single degree of freedom system with three typical nonlinear types(bilinear stiffness,cubic stiffness,quadratic damping)are firstly studied.The nonlinear pattern recognition is carried out by using the nonlinear feature indexes and the support vector machine method.Then,the parameters of the identified nonlinear model are identified based on the regularized least squares method.The dynamic characteristics and restoring force surfaces of a multi-degree-of-freedom nonlinear cantilever beam with three typical nonlinear types(bilinear stiffness,cubic stiffness,quadratic damping)are studied.The nonlinear pattern is identified by using the nonlinear feature indexes and the support vector machine method.Then,the parameters of the nonlinear model are identified based on the instantaneous feature-based and the regularized least squares method.Based on the Benchmark model proposed at IMAC XXXII(2014),the nonlinear element model validation and parameter identification are studied.In the benchmark model,the stiffness softening of the intermediate flexible structure under strong instantaneous excitation load is considered,which results in the nonlinear dynamic behavior of the structure.The nonlinear element pattern of the benchmark structure is effectively identified based on the proposed method.Furthermore,the parameters of the nonlinear model are also identified based on the instantaneous feature-based method and the regularized least squares method.