Pattern recognition approach to evaluation of liquefaction potential: exploration of Buller site

Abstract

Since November 1989, a research project aimed at applying the pattern recognition technique to the problem of estimating soil liquefaction potential has been in progress at the University of Canterbury. This report describes results of the first stage of the project, the acquisition of field data for the calibration of the pattern recognition model.

The aim of our research is not only to conduct an exploration of liquefaction sites in the Buller region, but also, more importantly, to develop a new approach to the evaluation of soil liquefaction potential. The new approach is based on the hypothesis that any layer of soil suffering seismic load which has similar characteristics (soil properties, environmental factors, earthquake characteristics) to soils at other sites which have experienced past earthquakes, will behave in the present earthquake as they behaved in the past. In other words, the new approach is a kind of field performance approach. The new approach needs to solve two aspects of the problem:

1. How to characterise efficiently soils that have suffered seismic load
2. How to perform a precise similarity comparison for soils in different states.

Our research was designed to solve the problem in the following two steps:

1. Setting up an information base which should contain as much information from historical liquefaction cases as possible;
2. Developing a recognition system based on the information obtained to provide a reliable liquefaction evaluation for the soils of interest.

This report describes the first step: the field exploration of liquefaction sites, to set up an information base which can be used to design a recognition system for the evaluation of soil liquefaction potential. In order to set up a reliable information base, we need not only to utilise all of the current information about the liquefaction, but also to conduct investigations at both liquefied and unliquefied sites in regions of different shaking intensity in historical earthquakes.