The British Geotechnical Association (BGA) has set up a small working group looking into the role of numerical methods in design. This group recently conducted a preliminary survey of practitioners to assess the industry's experiences and views. The general feeling was that Eurocode 7 (EC7) raises lots of issues without giving rm guidance on what to do, that this is leading to confusion, and is a subject that should be investigated by the BGA.
Of immediate concern is the mechanism by which existing numerical codes should implement the partial factor approach, for example, when using the strength reduction technique in ultimate limit state design. This is relevant in analysing the response of complex structures to a range of loading conditions, such as those involving staged construction and soil structure interaction, and in analysing the sensitivity of response to variations in key parameters.
There is uncertainty about how to apply load factors to structural elements and material factors to soil properties, given the inherent non-linearity of soil behaviour.
Another concern is selection of inputs for numerical codes. EC7 states that characteristic values of material properties should account for the variabilities of the property values. But it gives little guidance as to how this should be achieved, although it does suggest statistical methods as a possible way forward. While stochastic analysis, which is inevitably numerical, may be unlikely for routine design, the industry needs a range of studies to substantiate partial factors and understand how these factors relate to, for example, coefcients of variability found in the results of a site investigation.
High quality laboratory and in situ test data are paramount in the proper use of numerical models and ground investigations need to recognise the special needs of more sophisticated methods of analysis implied by EC7.
Certainly, if numerical modelling is worth doing, it is worth doing well.
Finite element and similar software should be driven by properly trained individuals. It is obviously preferable for these engineers to be part of a fulltime modelling group as experience is maintained across a variety of products and those doing the work develop a feel for the link between simulation and reality. The modelling engineer must give proper consideration to inputs and outputs, requiring appropriate experience in geology, soil mechanics and, possibly, structural engineering - given that there is often some soil structure link. But where are we to find these paragons of engineering virtue?
Of course this range of skills may not reside with one individual and a team approach may be needed. But the underlying philosophy should be that numerical modelling in geotechnical design is an activity for highly trained, albeit specialist, geotechnical engineers; that is, not an activity for 'detached' numerical analysts who have little interest in the materials being analysed. This has three implications.
First, industry needs to attach proper value to geotechnical engineers who have expertise in numerical modelling; second, universities need guidance as to what industry needs are in this area; third, industry needs to perhaps re-think the design process with all the attendant issues in quality assurance and the like.
When considering numerical modelling, it is important to recognise that methods that model deformation and failure mechanisms are better than simpler and more idealised calculation methods.
There will be a BGA meeting on the use of numerical methods in geotechnical design at the Institution of Civil Engineers on 26 February from 2pm to 4.30pm. Brian Simpson, Arup Geotechnics, will set the scene with an overview of EC7; Mike Jefferies, Golder Associates, and Tony O'Brien, Mott MacDonald, will share their experiences and views; nally, Helmut Schweiger, Graz University of Technology, will give a European perspective.