The collection of geomechanical information is a complex and dynamic process. As new information is available, the geotechnical model can be updated. However, this is not a straightforward process since the information sources may have different characteristics and reliability levels. This is a process based on judgment and experience. It lacks a systematic and mathematically valid process to deal with this problem and use new information to update the model parameters in order to reduce uncertainties. In this paper, it is intended to provide a contribution on this subject. A generic Bayesian framework for the updating of geomechanical parameters is presented. Emphasis is given to the theoretical aspects of uncertainty and to the Bayesian formulation. This framework is applied to the case of the deformability modulus updating in an underground structure. The prior distribution of the parameter was obtained through de application of analytical solutions based on the empirical classification systems. This distribution was then updated using the framework together with the results of a high quality in situ test. The Bayesian framework showed to be a mathematically valid way to deal with the problem of the geomechanical parameters updating and mostly in the uncertainty reduction related to the parameter real value.


In the construction of underground works several decisions are carried out under uncertainty. These uncertainties are related with two major problems, namely the geological/geotechnical conditions and questions related with the construction itself (advance rates, costs, etc). Figure 1 represents, in general terms, the decision cycle, adapted for engineering purposes, which is also applied to the underground structures construction [1]. In a first step, parameters are determined and included in engineering models. Then, based on their results, decisions are made with a given uncertainty degree. After new information is gathered the knowledge about the analyzed problem can be updated and reused in the models to obtain new results and perform decisions based on less uncertain data. The formal assignment of uncertainties and the updating procedure in order to improve the predictions are two critical aspects of this approach. The first has been already performed in many areas of geotechnical engineering like landslides [2, 3] and tunnelling. In this field, the "Decision Aids for Tunnelling" can be referred [4, 5, 6]. It is a procedure and a computer code which allows formalizing uncertainties related with geological and construction aspects.

Fig. 1. The decision cycle [1].(available in full paper)

However, only a few formal and mathematically consistent updating schemes have been developed in geotechnics. This is normally carried out using methodologies based on Bayes theorem. In [7] the author developed a Bayesian approach to update fracture characteristics with the results of flow tests. [8] used the observation of cracks in pavements to refine uncertainties concerning surface creep in slopes. Concerning the tunnelling field, [1] used the Bayesian framework together with a Markov process to update the mean length of the geotechnical parameter state (like "intense jointing"). [9] also used a Bayesian approach in order to update cost in tunnels construction. Concerning the geomechanical parameters updating it is not known any study to implement a formal updating framework.

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