Regional Seismic Damage Prediction Based on High-Performance GPU Computing: A Case Study of Tsinghua University Campus

Abstract

The prediction of seismic damage to buildings in urban areas is an important technique for the mitigation and prevention of earthquake-induced destruction. A methodology is proposed for such predictions, based on high-performance graphics processing unit (GPU) computing, which is applied to a case study of a real university campus. Specifically, an overall framework is designed based on cooperation between the GPU and CPU. A high-fidelity structural model suitable for GPU computing is built and the thread structure of the GPU-based seismic damage prediction is discussed. Taking Tsinghua University campus as an example, the seismic damage to the entire campus is predicted efficiently by the proposed method and the detail of local damage on different stories is obtained explicitly. Furthermore, an increase in speed of approximately 21 times is achieved in comparison with a traditional CPU approach. The outcome of this study provides a critically important reference for the prevention and mitigation of urban disaster.