ABSTRACT

This research proposes a surrogate model to predict the seismic response of individual structural elements in structures whose inherent vertical and horizontal irregularities result in components with different seismic vulnerabilities. A frequency-based data-driven model was developed which predominantly uses the frequency spectrum of earthquakes as input data. The seismic responses of several structural components can be simultaneously generated as output using the proposed model. A comparison of structure fragility assessments obtained with a conventional approach, and the proposed Deep Learning-based approach, was conducted to verify the accuracy of the proposed method’s prediction capability.

摘要

本研究提出了一个替代模型来预测结构中单个结构元件的地震响应，其固有的垂直和水平不规则性导致组件具有不同的地震脆弱性。开发了一种基于频率的数据驱动模型，该模型主要使用地震的频谱作为输入数据。使用所提出的模型可以同时生成几个结构组件的地震响应作为输出。对使用传统方法获得的结构脆弱性评估与所提出的基于深度学习的方法进行了比较，以验证所提出方法的预测能力的准确性。