High Performance Computing
National Cheng Kung University
With the dramatic increase of computer processing power in recent years, artificial intelligence, deep learning and big data technologies are becoming more mature. These computing advances are enabling scientists to solve problems that used to be unsolvable, but many complex scientific problems remain. One of these challenges is the accurate prediction of natural disasters.
Taiwan is located in one of Earth's biggest seismic zones, and as a result Taiwan has suffered severe damage from earthquakes over the years. NCKU's Professor En-jui Lee, who had previously participated in a 3D full-waveform seismic tomography project in southern California, created a high-performance computing platform for a 3D full-waveform seismic tomography system and earth sciences in Taiwan, on a limited budget.
In recent years, the processing power of GPUs has greatly improved. CPUs and GPUs apply different strategies for their processing tasks. CPUs, which can have around a dozen cores, are used to optimize sequence processing. GPUs, on the other hand, have thousands of cores, and excel at processing multiple tasks at the same time in parallel. In this age of rapid developments in artificial intelligence and deep learning, GPUs have become an effective tool for scientists to achieve results much more quickly.
Professor En-jui Lee noted that the objectives of 3D full-waveform seismic tomography can be divided into three aspects. Firstly, full wave field waveform analysis and simulations with multiple iterations are used to improve the resolution of current 3D velocity models. Secondly, strain tensors are calculated using stored high-resolution models, in order to improve the speed and accuracy of earthquake hypocenter simulations. Lastly, more accurate ground motion predictions are applied to improve the reliability of earthquake disaster analyses.