Imperial College recently organised a symposium to mark the 50th anniversary of its postgraduate course in soil mechanics. A key component of the event was a competition in which participants were asked to predict the performance of a single driven steel pile, an identical steel pile with a jet grouted bulb at its base, and as an optional component, the behaviour of the jet grouted pile under cyclic loading.
The problem was chosen because of Imperial College's involvement in an industry/ research collaboration to investigate through field experimentation this exact situation at the Zip des Huttes site near Dunkirk in Northern France.
Participants were given a comprehensive 34-page inform-ation pack with details on the ground conditions and as-constructed pile and jet grouting operations. Predictions were judged against Imperial's extensive field testing which was carried out under the on-going EU funded Gopal project.
Entrants were asked to provide a brief description of their calculation methods, the total load capacity of both piles, broken into base and shaft components, normalised load-settlement curves for both piles and optionally, the mode and direction of failure under cyclic loading, plus the number of cycles to failure.
Imperial was conscious of the need for impartiality and established an independent panel of judges under the stewardship of CIRIA's Fin Jardine. Under the rules of the competition, only the originators of the top three predictions were to be revealed, and the identity of the participants was not known even to the judges.
In total 16 predictions were received from seven countries. Nine entries were from consulting and four from contracting, but significantly none were from academia. Results are summarised in the four figures, which also show Imperial's field measurements.
There was enormous scatter in the predictions and no one approach or method consistently out performed the others. The wide scatter was not just for the jet grouted pile (a novel technique, with no previous experience), but also for the control pile (a common problem in foundation engineering).
So few attempted to predict the dynamic behaviour, and those who did achieved such poor results, that the judges decided to remove this element from the competition.
For those in the know, the wide scatter of results was unsurprising and conforms closely to the observation that industry standard methods of predicting the capacity of a single pile are reliable only to about plus or minus 60%.
Too often it is assumed that the behaviour of a single pile is one area of geotechnics that is well understood, well documented and can be taken from the basic textbooks with confidence. The Imperial predictions competition may prompt the industry to focus on the big unknown in the equation - construction effects on pile behaviour. There is clearly a long way to go in geotechnical engineering to close the loop on theory, design and construction practice.
The competition poses the question, if predicting pile behaviour is so difficult, why bother paying for geotechnical expertise?
First, the best predictions were reasonable, underlying the need for good geotechnics, as opposed to second-rate geotechnics, or geotechnical design carried out by structural engineers.
Second, good geotechnical engineers know the limits of accepted theory. Armed with a knowledge of the limitations of current design practice, it would be relatively straightforward and certainly cost effective in a real construction project, to have 'calibrated' the pile design using on-site static load tests on trial piles before the piling contract.
As Professor Peter Vaughan commented at one point during the Imperial Soil Mechanics' 50th anniversary celebrations, 'I have often pondered on whether our subject is extraordinarily difficult, or whether it just attracts thickies.'
Knowing the former to be the case, the challenge in the immediate short term must be as much to do with understanding our limitations, reducing the risk when engineering with these unknowns, and articulating to geotechnical clients the potential risks involved when engineering in soils.
Judges' preliminary observations
Few contestants gave sufficient information to understand the methods used. Five assumed the pile plugged, three assumed it was unplugged and eight did not specify.
For load capacity predictions participants used the following techniques:
French and Dutch codes
Jardine & Chow's method
Load displacement predictions were based on t-z curves, hyperbolic functions, advanced numerical analysis (Plaxis & Flac), 'eye-balled' in.
There was an enormous scatter in prediction and no one method consistently out-performed the others.