ABSTRACT

The spatial interpolation of geotechnical information is one of the most representative applications of geo-informatics database. It is necessary to estimate the soil properties at arbitrary position in the ground from the data of soil investigation. In this study, an artificial neural network was applied to interpolate spatially the soil properties, such as natural water content, liquid limit and one-dimensional compression curve. The accuracy of spatial interpolation of soil properties was judged based on the four indexes, such as R2, G, MARE and SR. The artificial neural network estimates the appropriate soil properties with high accuracy, so that the availability of spatial interpolation of soil properties through artificial neural network was confirmed.

INTRODUCTION

There are large amounts of geotechnical information obtained from soil investigations in many construction projects in Kansai area and they have been accumulated into geo-informatics database, called GIbase (Tanaka & Tsukada, 2009). It is very effective to handle the practical problems in geotechnical engineering, such as construction of buildings, development of underground spaces as well as geo-environments and disaster presentations. Therefore, it has been widely applied in many geotechnical fields. The estimation of geotechnical information at sites in which the soil investigation have not been carried out is one of the most significant applications of geo-informatics database. A technique to interpolate some revealed geotechnical information spatially must be required in order to estimate geotechnical information at an arbitrary position. There are many mathematical techniques to interpolate spatially information. The practical and useful methods for interpolating geotechnical information are not yet established. By the way, an artificial neural network is one of the information processing systems which have been advanced recently. This technique has been applied to data mining. Data mining is the process of discovering new patterns from large number of data sets. That is, locations at which soil investigations were carried out and geotechnical information obtained from soil investigations can be chosen as data sets.

This content is only available via PDF.
You can access this article if you purchase or spend a download.