ABSTRACT:

The use of numerical models requires determination of input parameters. Backanalysis of some test or trial construction observations must be performed. This process usually produces a non-unique set of parameters. The effect of parameter variation must also be examined. The focus of judgement is on the model parameter values. The method of predictive analysis is a valuable technique that enhances backanalysis and numerical model predictions. The method forces the analyst to define key events for which the prediction of worst outcomes is required. Predictive analysis determines the set of model parameters that minimises (or maximises) these key events whilst maintaining an acceptable fit of the model predictions with the observations. The method is applied to a problem of estimating tunnel settlements using numerical modelling and pressuremeter test results. The example demonstrates the process and highlights the point that the focus of engineering judgement remains on the model performance with respect to the observations rather than the model parameters themselves.

INTRODUCTION

The use of numerical modelling is now widespread in geotechnical engineering. The designer can examine new construction methods in ground for which no database of previous construction exists. A sensitivity analysis may be carried out on key input parameters to estimate the likely range of outcomes. Even so, the most important issue is the choice of values for the model input parameters. Nearly all models used in geotechnical analysis require a value of Young's modulus. There are numerous methods for estimating a value. These include: direct tests on pieces of intact core, correlations with RQD, RMR and Q values, use of seismic velocities, and the GSI method of Hoek et al (1995). Engineers usually make a number of estimates for Young's modulus, which are likely to vary from lowest to highest by as much as a factor of 20.

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