Prediction of the debris flow using the combination of the neural networks and GMDH.

点击翻译按钮获取中文摘要

We develop the prediction system for debris flow using a combination of the networks and the GMDH.1) Make an archive of precipitation radarWe collect the radar data from 1982 to 1985, 1993 to 1997 observed with Kyushu Nanbu radar installed by Ministry of Constrution. We also collect the rain gauge data during the same time observed with AMEDAS installed by Japan Meteology Agency. All collated data are saves with CD-ROM and made a database using java. This database is possible to search the data through the Internet. The URL of the database is http://www.radar.civil.ac/2) Development of the Prediction systemWe ported a GMDH program from Turbo PASCAL to Java. This ported program is possible to use as applet or application and get teh precipitation data from the database server through the Internet.3) Future workWe are planning to develop the prediction system using GMDH. A training program is select the most good parameter using GMDH before the rainfall. When rainfall starts, a prediction program get the parameters from the training program and real-time precipitation data through the internet and predict the debris flow.