The authors' contribution to the complex problem of design for piled embankments on poor ground is relevant given the prevalent uncertainty which the authors draw attention to. Two case histories were investigated - the A13 and the small Second Severn Crossing (SSC) trial patch test (Bell et al). Since details of the actual full-scale works for SSC including records of actual settlements were published (see references 2 and 12 of the paper) it is difficult to understand why, in preference, the authors focused on the small-scale pre-contract trial at the site.
When deciding upon the validity of various design approaches, proven construction work must surely provide the yardstick for judgment. Clear comparison between the design prediction and actual observation of full- scale constructed works is a prerequisite for any meaningful consideration of comparison between methods. In the present case the authors have compared predictions based on their model rather than with observed results. Consequently the authors conclude that the 'only reliable method of design is by three dimensional numerical analysis'. However this method would have considerably overpredicted the reinforcement loads required in the SSC toll plaza project, and several other projects designed and successfully constructed around the world.
The data on SSC available in references 2 and 12 of the paper gives a large amount of information on the actual deformations, pressures etc. Any numerical model requires calibration, and predicted deformations from the authors model appear to be at least a factor of 10 above the deformations measured in the field. Therefore, at least for SSC, the authors proposed method was unable to approximate to the observed behaviour.
The design of the SSC work was based on basic geotechnical principles combined with the so-called 'Guido Method' to generate arching to transfer the embankment loading on to the VCC heads. This design concept was subjected to 2D finite element axisymmetric and plane strain analysis, Plaxis 4.10, by Halcrow-SEEE and also external Category 3 checks. The fundamental principle of the SSC design philosophy is that relatively close trangular grid would support the full vertical load, plus HB surcharge, wholly by the safe capacity of the VCCs. The specified post-construction settlement criteria would be achieved by the 'gothic arch' principle where the compacted rock fill would arch at 45 to ensure post-construction differential measurements at road surface of less than 1 in 1,000. The prime reason for the geogrid was to promote and ensure that this arching angle would be immediately developed from the VCC head. Strictly speaking, once this arch is locked in place there is no need for any geogrid to support the fill beneath it. However, it was considered prudent to adopt sufficient strength of geogrid to support the weight of the trapezoid of non-arch fill.
The reason for the trial was therefore to confirm that:
a) the anticipated arch would in fact develop and remain intact even with large deflection of the geogrid.
b) the geogrid was capable of supporting the fill below this arch.
The geometry of the square grid layout of the VCCs was deliberately chosen, in comparison to the proposed design of triangular grids, to generate much larger deflection of the geogrid than would be expected in the actual works. The steel kentledge blocks were added to attempt to push the trial to the limit. The Maddison et al paper (reference 12) clearly states that the trial embankment system limited the extent of the arch forming above the transfer platform and it was therefore a more severe test than that which would occur in practice.
The trial confirmed that all basic design assumptions of the composite system were valid and the contract proceeded as per the original design.
With regard to specific comments and assumptions made by the paper's authors, the installation of the 17 No settlement rods at various levels to measure deflections severely inhibited the compaction of the trial embankment fill. In fact, there was no compaction to the top 0.5m. The Tensar SS2 layers were also placed orthogonally in order to provide similar stiffness in both directions.
We also note that the paper's authors have used the much lower 120 year reinforcement stiffness to model deflection during construction of the embankment. This is clearly incorrect. However, the Guido method should not be modelled by a single layer of reinforcement at the base of the load transfer platform. The whole point of including the grid reinforcement to different levels is to have an integral restraining function applied to the granular fill which with a carefully chosen granular fill of p 45 will interlock with this material. Simply modelling the integrated reinforcement as cable elements at the base of the layer does not adequately reflect the mechanism that develops.
The paper's authors state that it is easy to modify their method to take account of triangular pile layout. In the event that they wish to re-model and recalibrate to analyse the as-built SSC results, we would point out that the typical embankment settlement at SSC was less than 20mm, the additional settlement of Maddison at al figure 14 location 3 being due to the presence of a loose to medium dense sand layer about 1.0m beneath the toe of the VCCs and that the 12 month monitoring of instrumentation and ongoing level surveys have, we understand, revealed negligible further movements. These settlements are in direct contrast and of a totally different order to those predicted in the authors' paper.
In conclusion, in our view the authors are correct in that, by development of arching from the pile/VCC head, large deflections at reinforcement level will not necessarily lead to large differential movements at the surface. However their approach does not appear to correctly model the observed behaviour of actual structures on the Toll Plaza Project for the Second Severn Crossing. Further development of the model, calibrated against further full scale performance, may prove fruitful.