Development of Evaluation System for Regional Seismic Input Motions Utilizing Neural Network Data Processing

点击翻译按钮获取中文摘要

The purpose of this study is to develop a new evaluation system for seismic input motions in urban area utilizing a data processing technique of neural network. Sapporo metropolitan area was taken as a test field and a prototype model was constructed.At first, neural network data processing techniques were reviewed trough many references from various fields and discussed a way to apply them to seismic input evaluation problems. A prototype model for evaluation of regional distribution of seismic severity was constructed on PC with an ordinal method of analytical approach paying an attention to the future extention. In parallel, seismic intensity data and strong motion records in the area were accumulated. In 1993, during the research period, we had two large damaging earthquakes in Hokkaido and detailed distribution maps of seismic intensities in the area were made through questionnaire surveys. And strong motion records were obtained at two sites by newly installed seismometers. Using these data as teacher's ones, inferring models were constructed adopting neural network data processing and improving the proposed evaluation mopdel.