Generalized neural network based model for structural dynamicidentification, analytical and experimental studies

摘要

The paper presents simulation and experimental studies ofidentification of civil engineering structures using neural networks.The identification of structural models by using measured data is animportant issue in engineering. Although static function mapping may beachieved using neural networks without knowing the fundamental physicsof the system, dynamic model identification is still a challenging topicin neural network applications. A generalized neural network-basedtechnique for structural dynamic model identification is developed basedon the dynamics of structure. During the simulation study, structuralresponse records from a 10-storey San Jose apartment building subjectedto three different earthquakes are adopted for the dynamic modelidentification. For the experimental study, a series of experiments wereconducted in which a designed scaled model structure, mounted on a shaketable, was tested. The neural network is trained and examined using themeasured structural responses under different earthquake loadingconditions. It is shown that the trained neural network is capable ofproviding sensible outputs when presented with input data that has neverbeen used during its training