TEACHING OF geotechnical engineering at university is in need of a major overhaul, Imperial College professor of analytical soil mechanics David Potts said last week.
Delivering the prestige British Geotechnical Association Rankine Lecture, Potts stressed that numerical analysis is 'by far the best tool at our disposal', and urged that young engineers need to be far more familiar with constitutive modelling and modelling software.
'This has to be taught at undergraduate level, ' said Potts.
'But it would overload students, so what do we teach less of?
Peripheral subjects like management? Or do we replace conventional analysis?'
In arguing both for and against the motion: 'Numerical methods of analysis have reached the stage where they are superior to conventional approaches and can replace them in the geotechnical design process, ' Potts warned that despite its advantages, there were many pitfalls when using numerical analysis which engineers need to be aware of.
Using a number of wellknown projects, he described how numerical analysis was 'in a league of its own' when modelling soil-structure interaction and providing mechanisms for behaviour.
He explained how numerical analysis showed the Pisa's Leaning Tower was due to leaning instability rather than bearing capacity failure. 'If it had been bearing capacity failure, then loading up with lead weights [to stabilise the tower] would have been suicidal, ' he said.
He also argued that the complex soil-structure interactions for the Westminster Station box on the Jubilee Line Extension could only have been modelled using numerical methods. 'Try modelling this in a centrifuge, ' he said.
But numerical software often contained 'bodges and fudges, often disguised by technical jargon, ' he warned. It also gave non-sensible predictions and was difficult to use for simple problems, Potts said, 'giving less conservative results'.
But the most worrying problem was user dependence.
Recent benchmarking exercises showed it was impossible for even experienced engineers to obtain the same results. In one case, 12 teams investigating the behaviour of a deep excavation retaining wall came up with widely different results.
'As five teams used the same software, the software can't be blamed. There was a significant amount of human error.
'In the future, numerical analysis will play a pivotal role in design; the pitfalls are many but not insurmountable.'
Everyone had a part to play, he said - academia, industry, institutions, and software suppliers.
He was convinced that this would 'produce safer, more economic and more effective geotechnical structures'.
Pott's paper will be published in Géotechnique.
Next year's Rankine Lecturer will be Mark Randolph from the University of Western Australia.