A Critical Review of the Observational Method

University dissertation from Stockholm : KTH Royal Institute of Technology

Abstract: Building a sustainable structure in soil or rock that satisfies all predefined technical requirements implies choosing a rational and effective construction method. An important aspect is how the performance of the structure is verified. For cases when the geotechnical behaviour is hard to predict, the existing design code for geotechnical structures, Eurocode 7, suggests the so-called “observational method” to verify that the performance is acceptable. The basic principle of the method is to accept predefined changes in the design during construction, in order to accommodate the actual ground conditions, if the current design is found unsuitable. Even though this in theory should ensure an effective design solution, formal application of the observational method is rare. It is therefore not clear which prerequisites and circumstances that must be present for the observational method to be applicable and be the more suitable method.This licentiate thesis gives a critical review of the observational method, based on, and therefore limited by, the outcome of the performed case studies. The aim is to identify and highlight the crucial aspects that make the observational method difficult to apply, thereby providing a basis for research towards a more applicable definition of the method. The main topics of discussion are (1) the apparent contradiction between the preference for advanced probabilistic calculation methods to solve complex design problems and sound, qualitative engineering judgement, (2) the limitations of measurement data in assessing the safety of a structure, (3) the fact that currently, no safety margin is required for the completed structure when the observational method is applied, and (4) the rigidity of the current definition of the observational method and the implications of deviations from its principles.Based on the review, it is argued that the observational method can be improved by linking it to a probabilistic framework. To be applicable, the method should be supported by guidelines that explain and exemplify how to make the best use of it. The engineering judgement is however not lost; no matter how elaborate probabilistic methods are used, sound judgement is still needed to define the problem correctly. How to define such a probabilistic framework is an urgent topic for future research, because this also addresses the concerns regarding safety that is raised in the other topics of discussion.