At a joint November meeting of the British Tunnelling Society (BTS) and the British Geotechnical Association (BGA) on the Nicoll Highway Collapse, in Singapore, BTS chairman Bill Grose introduced Richard Davies, a director and geotechnical consultant of Benaim. Davies first became involved in excavation problems in the difficult ground conditions of Singapore in 1971, during the construction of one of the first basements for high-rise buildings, and subsequently with many other excavation problems, including all phases of the Mass Rapid Transit System (MRTS). Following the Nicoll Highway collapse, he was appointed by Singapore’s Land Transport Authority (LTA) to investigate the cause of the collapse and report to the Committee of Inquiry (CoI).

Davies began the presentation stating four people had died in the collapse on 20 April 2004. The LTA also appointed three other experts: Prof Andrew Whittle (MIT), Brian Bell and Prof Chiew Sing Ping (NTU). Evidence of fact was given by 172 witnesses and 19 experts appointed by five parties. Hearings were held over eight months from August 2004 to March 2005.

Davies described the extent of Contract 824 on the new Circle Line on the MRTS and the site plan (figure 1). The area of collapse was the cut & cover works to the west and adjacent to the temporary shaft (TSA Shaft) that was 20m wide and planned to be 33m deep. The construction sequence is shown in figure 2.

The collapse occurred following excavation below the level 9 struts and removal of the upper jet grouted layer.

Davies explained that this was probably one of the best documented collapses ever, as there was a great deal of data available. In the area of the collapse, inclinometers had been installed in the diaphragm wall on the north side and just behind the wall on the south side. Strut loads were also monitored with pressure cells at the upper levels and strain gauges at the lower levels. In addition, pore water pressures and settlements were also being measured.

The layout plan at level 9 struts is shown in figure 3, it also shows the position of the 66kV cable utility bridge over the excavation and the gaps in the diaphragm walls that were filled with a combination of sheet piling and grouting; the position of inclinometers I-65 and I-104 on the north and south sides respectively; and the instrumented strut S335 (coloured red). The gas main that also crossed the excavation is located just outside the plan to the west.

The collapse occurred at approximately 3.30pm on 20 April 2004. It was fortunate that the traffic lights further up the highway had only just turned green at the time and the traffic had time to stop before reaching the collapse area.

The instrumented Section S335 (figure 4) shows the geology with the Old Alluvium considerably higher on the north side (the Nicoll Highway side) and the resulting shorter length of diaphragm walling embedding on the south side and the location of the inclinometers. The field measurements were illustrated as a series of deflection plots. The deflections with the excavation below strut level 7 were indicating a smooth profile on the north side where the inclinometer is in the diaphragm wall, as well as a degree of fixity provided by the Old Alluvium. In contrast, the deflections on the south side were bumpy and larger than the north side and a rotation from the bottom of the wall, indicating a lack of fixity.

Warning sign

There had been a number of ‘thong’ sounds from failures in the waling beam stiffeners. The strut S338 at level 9 had failed between 9.00am and 9.15am on 20 April at the strut-waler connection where a sway movement had been recorded in the C-channel that had been used in place of plate stiffeners.

Probably the most important single set of measurements was the hourly strut loads at S335 at levels 8 and 9 (figure 5). The loads in the struts were not considered to be accurate but nevertheless the changes are considered to provide a good indication of how the loads were changing. The last series of inclinometer (figure 6) measurements had shown a further dramatic increase in deflections; the excavation of the upper jet grouted layer had only started in the evening after the deflection measurements had been taken on 17 April and the last measurement at 1.00pm on 20 April at the south side showing a further increase in the deflection to over 400mm. The deflections had generally exceeded predictions and had been back analysed and an increased prediction calculated as illustrated in figure 7.

The last information available before the collapse, from eye-witness accounts, was the ‘thong’ from the level 8 waling that was also observed to buckle, following which the excavation area was evacuated.

Davies said that it was not possible to know exactly the scenario of the failure but it is thought that the struts had failed both vertically and laterally as illustrated by figure 8. The factors involved with the collapse were given as the ground conditions, the site history, the wall analysis and the strutting system.

Davies described the geology of Singapore with the erosion of the rocks and the infilling with very soft materials that had resulted in complex surfaces particularly with the Old Alluvium. A contour at the base of the very soft marine clay across the site of the collapse indicates a sharp drop of 10m from the north to south sides and the presence of a buried channel. Boreholes shown at this section illustrated the complexity of the materials. In addition, there were the potential effects of the reclamations in 1930-1940’s and 1970’s of the site from the Kallang Basin on the consolidation of the marine clays at depth, which may not have been fully appreciated in the analyses of the design of the walls.

Davies then referred to some piezocone tests carried out in this area which was in the later reclamation. The conversion into undrained shear strengths using an Nk of 14, which was considered appropriate for the excavation problem, indicated that there was practically no increase in shear strength below the stiffer layer that separated the upper and lower marine clays when correction for the overburden pressure is taken into account.

Davies then discussed the design processes that had been followed from values taken from the geotechnical report and the difficulties with the design programme used, which had two methods of analyses that gave either effective strength parameters (Method A) or total stress analyses. He demonstrated that Method A had overestimated the undrained shear strength by some 50% and underestimated the bending moments by a factor of two.

Davies also discussed the analyses of the jet grouted layers assuming them to be an elastic plastic material. For the strutting system, he explained that for the upper levels single struts were used and some were bearing directly onto the walls, but the majority were double struts with splays going onto waling beams. Down to level 6, plates were used to stiffen the web at the waling beams. In actual construction, the stiffeners below level 7 had been changed to C-channel ones. This was further complicated by the curved structure that made it physically difficult to fit the struts and some splays were omitted (as can be seen in figure 3).

Unfortunately the strut-waler connection was not revisited and the removal of the splay resulted in an increase load of some 30%, and in addition he explained that if BS 5950 is used for the design of the strut-waler connection, this connection was shown to be the weakest section with only half of the required capacity.

The difference in the behaviour of the plate and C-channel stiffeners was examined by some work by Prof Chiew. A Finite Element (FE) analysis of the plate stiffener connection demonstrated the buckling of the plate and bulging of the flange of the waler. A similar FE analysis for the C-channel was very different with a sag to the whole waler beam which was similar to that seen on site. The capacity of the connection was similar but demonstrated a more brittle behaviour. These results were verified by physical tests and even the ‘thong’ sounds were replicated.

Referring again to the strut loads (figure 5), Davies said it was believed that the loads at level 9 had not increased as the yield point had already been reached early on 20 April. Further, the load for the level 8 struts had been designed for the maximum load on the strut on the way up and the capacity based on BS 5950, which was going to be conservative and was 50% greater than that at level 9. Hence, the level 8 struts were able to carry the additional load being shed from level 9 until 3.00pm on 20 April when a ‘thong’ was heard and failure observed.

Further verification of the failure mode was demonstrated by modelling by Prof Whittle using the actual depths of the Old Alluvium, considering the walls as elastic plastic and the struts also treated as elastic plastic, but with the struts to level 6 being ductile and the lower struts as more brittle.

Davies summarised the main issues as being: the buried channel, probable under-consolidation of the lower marine clay, over-prediction of the undrained shear strength in the original FE analysis, the consequences, properties and behaviour of jet grouted layers, load at the strut-waler connection and splays omitted, design of strut-waler connection to BS 5950 and the brittle behaviour of the C-channel stiffeners and factors involved. The CoI concluded the collapse was rooted in two critical design errors:

  • Under-design of the diaphragm wall associated with the use of a soil simulation model that over-predicted the undrained shear strength of the clay.
  • Under-design of the strut-waler connection associated with the over-estimate of capacity based on BS 5950 and splays omitted.

The CoI also commented that the change from plate stiffeners to the C-channel was a major contributing factor. If only the plate stiffeners had been used throughout, the failure would probably have been localised and slower.

Davies concluded by referring to some of the consequences. These included the abandonment of the previous Nicoll Highway Station, the cut & cover works and the bored tunnels under Kallang Basin. The station was moved south to a new location and the bored tunnels taken on a revised alignment right up to the station. The diaphragm walls for the station were increased to 1.5m thick, inclinometers placed at frequent intervals around the perimeter, top-down construction used, walls were embedded into the Old Alluvium and 8m of jet grouting used. There were knock-on consequences for other LTA projects and now two and a half years on work on the station had only just recommenced.

Inevitable?

The chairman then introduced Dr David Hight of Geotechnical Consulting Group who had been appointed as an expert by the contractor, Nishimatsu Construction, to present its’ case, ‘Was the Nicoll Highway Collapse an inevitable consequence of the design errors?’. The view expressed at the CoI was that the Nicoll Highway Collapse was an inevitable consequence of design errors, which led to the applied loads on the struts increasing with time and equalling the capacity of the strut-waler connection.

Hight suggested that this view ignored the uncertainties in predicting the applied loads at the time of the collapse and in the capacity of the strut-waler connection, the trends in the monitoring data, the timing of the collapse and the observed deformation of the strut-waler connection involving downward failure at both ends and its ductile-brittle response.

The capacity of the strut-waler connection was based on physical load tests. The predicted applied load was based on FE analyses, that in most cases treated the jet grouted layers as a ductile material, did not model the presence of the bored piles, or modelled them as walls in 2D analyses. It failed to set limits to the bending moment capacity of the wall or to allow for reductions in wall stiffness as cracking developed, and the measured loads were much lower than predicted.

The observed trends in the strut loads at levels 8 and 9 are shown in figure 9. As Davies had said, it was agreed by all parties that the trends in the strut load monitoring data were correct but not their magnitudes. The trends were consistent with there being yielding of the level 9 strut-waler connection when the excavation passed beneath but with no further significant changes in load in either the level 8 or 9 struts until the collapse initiated. These trends in load, and in wall deflections at all stages, were matched in FE analyses which modelled the jet grouted layers as a brittle material, took into account the effects of cracking and the formation of a plastic hinge.

Hight then referred to the buried valley in the Old Alluvium explaining that the excavation had passed beyond it, where strut loads would have been at a maximum, and was taking place in the vicinity of the 66kV crossing, the presence of which had necessitated changes to the bored pile layout. There was no evidence of load increases after excavation to the level 9 or of significant compression of the yielding connection at the level 9, which would have taken it beyond its ductile range. A trigger for the collapse was necessary.

A potential trigger was provided by the relative vertical displacement between the kingposts and diaphragm wall panels which caused a forced sway failure of the strut-waler connection. Figure 10 illustrates the effect of brittleness of the connection on time to collapse and the effects of a forced and unforced sway.

The forced sway can explain the timing of the collapse, the form of the observed distortions, the trends in the monitoring data, and the speed at which the collapse developed.


Plan map showing the area of the collapse Figure 1 – Plan of collapse The construction sequence of the box Figure 2 – The construction sequence Layout plan of the level 9 struts Figure 3 – Layout plan Cross section of Section S335 Figure 4 – Section S335 Measured strut loads at S335 Figure 5 – Measured strut loads The Last inclinometer measurements before the collapse Figure 6 – Last inclinometer measurements Back analysis Figure 7 – Back analysis Section S335 Figure 8 – Section S335 Questions & Answers

Chris Raison (Raison Foster Associates) asked if the correct undrained shear strength had been selected, did the speaker believe the problem with the waling connections would have caused the collapse?
Davies replied that if the design had used the original GIM shear strength using conventional undrained analysis and if BS 8004, which looks at factors of safety, had been applied, there would have been a much more robust design that could have tolerated a failure at the strut-waler connection at level 9. The diaphragm walls would have been thicker with greater embedment into the Old Alluvium and the struts at level 8 designed for a much higher load.
Ather Sharif (Civil Engineering Dynamics) was astounded at the lack of monitoring for a structure on a project with such a scale and an apparent lack of confidence in the observations. His point was that monitoring of structures requires a lot of experience and compared this to the aircraft industry where pilots are trained to deal with false alarms, etc, and the need to have other means of getting the relevant data. There seemed to be a lack of focus and belief in the monitoring.
Davies said that there had been a huge amount of measurement data. There was a difference between taking and believing measurements. All struts had been pre-loaded and the effects of the pre-loading could have been monitored. There were a lot of issues about measurement: these need to be continuous and there needs to be a proper process for responses.
Terry Crabb (London Bridge Associates) referred to the apparent lack of belief in the deflection movements.
Davies replied that this had been the only place where large movements had occurred. The strut loads had not exceeded their design values.
Nick Walsh (Tony Gee & Associates) noted that the analysis of the collapse did not take account of the jet grouted layers. There is a history of jet grouting not performing as designed and asked if there had been any assessment of the performance of the jet grouted layers.
Davies did not know, but it had not been possible to prove. The upper layer had only been 1.5m thick. The construction method had not been complicated. The collapse had contributed to jet grouting getting a bad name. The Chairman added that it was fair to say that the jet grout layers had resulted in discussion over several days at the Inquiry.
Duncan Nicholson (Arup), a member of the CoI, wondered if Davies would say anything about the back analysis process 1 and 2, the fact some 70% of the 31 inclinometers over the 1.5km length of excavation had exceeded their predicted values and the change in process of the removal of the jet grout layer over the whole length rather than removal in bays.
For the back analysis, Davies said that this had not been a proper process as they had only tweaked parameters and that they did not have a satisfactory model to start with. As far as the installation of the struts was concerned, there were practical problems and whether their installation would have averted the collapse is debatable. It is believed that there was a rapid failure of the strut-waler connection at S338, about 12m nearer the shaft and we know what went on at the instrumented section at S335.
KC Law (Coffey) said one of the statements was that Method A was an over estimate of the shear strength but with the deflection shape could this not have been an over estimate of the performance of the jet grouting.
Davies said that the difficulty was assessing what greater shear strength could be used. The other data on the shear strength were cone and vane tests. Davies doubted that the shear strength could be approaching a C/P value of 0.33. Also no allowance had been made for any contribution from the bored piles. In response to a supplementary question, Davies said that there had been no evidence that the workmanship was particularly bad.
Robin Saunders (Jacobs) asked whether the pre-collapse or post-collapse site investigation data had been used.
Davies replied that all SI data he had used was pre-collapse data.
Rapporteur – Anthony Umney