Technical paper: Productivity analysis of diaphragm wall construction for Prince Edward Road Station on Circle Line 6

Mark Smith, Land Transport Authority, Singapore

A diaphragm wall is a type of earth retaining structure typically used for basements, tunnels and deep excavations. It is a reinforced concrete wall that is cast in sections (or panels) within an excavated trench. The trench is temporarily supported by a stabilising fluid during the excavation process, which is then displaced during the placement of concrete.

As a foundation element, construction of the diaphragm wall is often a programme-critical activity, preceding excavation of the basement or construction of the superstructure. It is also an activity with high preliminary costs due to the large specialist plant required during construction. A better understanding of diaphragm wall productivity can therefore offer both programme and cost benefits.

This paper provides an empirical analysis of the diaphragm wall construction records from Prince Edward Road Station (PER) in Singapore, with two objectives:

1) Assess the effect of panel geometry (length, thickness, depth) on productivity to provide improved planning tools for diaphragm wall construction; and,

2) Reduce waste by identifying the variables that contribute to increased panel overbreak.

Overbreak is the term used to describe the caving of loosened material along the edge of an excavation. During diaphragm wall construction, the volume of over-excavated material is replaced with concrete during the casting process. This excess concrete is not required in the structural design and can be considered redundant. While not a direct measure of productivity, overbreak should be considered a wasteful and unproductive activity – incurring additional time and cost to dispose of the over-excavated material, and again when replaced with the equivalent volume of concrete.

To allow a reliable comparison of the data, all construction records are from the same project, undertaken by a single specialist contractor using the same construction method, and with similar ground conditions throughout.

Circle line 6 (CCL6) is a fully underground Mass Rapid Transit (MRT) system currently under construction by the Land Transport Authority of Singapore (LTA). It comprises three stations – Keppel Station, Cantonment Station and Prince Edward Road Station – with a route length of approximately 4km. Once completed, CCL6 will close the Circle Line loop by connecting HarbourFront Station and Marina Bay Station (Figure 1).

Prince Edward Road is a three-level underground station measuring 297m in length, up to 48m wide, and 28m deep. It is located within Singapore’s central business district, near high-rise properties and arterial roads. The site is also bounded by several historically important buildings that are potentially sensitive to ground movement.

Figure 2 shows the geological map for Prince Edward Road Station; a summary of these ground conditions is contained in Table 1. The site investigation identified 1m to 10m of fill material, overlaying Kallang Formation up to 12m thick (Estuarine Clay, Fluvial Clay, Marine Clay), including some localised pockets of sand. Beneath this lies the Jurong Formation, a variety of sedimentary rocks underlying most western and south-western areas of Singapore. These deposits are often faulted and fissured and can frequently exhibit a folder structure1. Varying weathering grades were identified at the site, ranging from moderately weathered (SIII) to residual soil (SVI). The depth of the rockhead (SIII) varies considerably across the station footprint but was found from depths of 15m below ground level.

The bore log information is consistent with the conditions experienced during diaphragm wall excavation, with highly variable and irregular rock conditions encountered throughout the site.

For this study, the construction process for each diaphragm wall panel has been simplified into four stages:

The prior construction of guide-walls has not been considered, as it can be undertaken independently and is typically not a programme-critical activity.

For Prince Edward Road Station, mechanical and hydraulic grabs were used to excavate the soft ground (Fill and Kallang Formation), and reverse circulation trench cutters were used to excavate the rock mass below. All excavation tools (grabs and cutters) were 2.8m in length and sized to match the diaphragm wall thickness, e.g. tool widths ranged from 0.8-1.5m. Where necessary, chisels were used to remove hard ground and to trim the excavation profile.

The panel layout was developed by the diaphragm wall specialist with the trench stability analysis undertaken to DIN 41262. The project specification limited the maximum panel length to 6.0m, with 3.0m panels specified near to sensitive structures.

Steel stop-ends were temporarily installed between all panels to a depth of 3m below the station base slab. Stop-ends were used to improve the joint quality and facilitate the installation of water-stops between panels.

As diaphragm wall works are considered to be a safety critical activity, they were undertaken on a 24-hour schedule, and the construction process was continuous once excavation had commenced. The quality of the diaphragm walls adhered to the requirements stipulated in the LTA’s Material and Workmanship Specification.3

The construction process for barrette piles differs slightly from that of a diaphragm wall panel, therefore the construction records for barrette piles are not included in this study. Similarly, irregular-shaped diaphragm wall panels used to form corners and intersections have been excluded from the study as these are known to have increased overbreak4.

When measured on plan, Prince Edward Road Station has 1,587 linear metres of diaphragm wall and barrette piles. Allowing for the above exclusions, this provided a sample size of 274 panels, with the following data ranges:

Construction records were produced by the specialist and verified by the main contractor and resident engineer.

The productivity rates have been compiled for the four diaphragm wall construction stages: trench excavation, de-sanding of bentonite slurry, installation of reinforcement, and placement of concrete.

Trench excavation is considered to be the critical activity in the diaphragm wall construction process5. For programming and resource planning, productivity is typically measured in linear metres per day – i.e. the length of wall, measured on plan, that each rig can excavate within a 24-hour period. As this is common practice within the local industry, the following assessment uses linear meters per day (m/day) to measure productivity. Panel excavation was undertaken continuously, therefore any unproductive time, such as rig breakdown or tool changes, is captured in the data.

To assess the effect of panel thickness on excavation duration, production rates were compared for panels of similar properties but varying wall thickness. Two sample groups were identified: 0.8m, 1.0m and 1.2m thick panels of depths 20-25m (sample size 55 panels), and 1.2m and 1.5m thick panels of depths 45-50m (sample size 144 panels). The excavation rate for similar panels of varying wall thickness is shown in Figures 3 and 4.

In both sample groups, the data does not support a strong correlation between diaphragm wall thickness and the excavation rate. While some individual construction records and anecdotal evidence suggests that increasing panel thickness may reduce the excavation rate, it is deemed not to have a significant effect in this data set. This may be because the excavation tools were sized appropriately to match the wall thickness, i.e. a 1.5m wide grab/cutter was used to excavate a 1.5m thick trench.

By assuming that panel thickness has an insignificant impact on excavation rate, the relative effect of panel length and depth may be compared. Table 2 shows the average trench excavation rates categorised by panel length and depth; the table also shows the sample size of each group.

The data shows a general trend of excavation rates decreasing with increased panel depth; however, the change is most significant between shallow panels (20-25m) and panels over 37m deep. Average excavation rates dropped from 1.5-2.6m/day for panels 20-25m in depth, to 0.3-0.7m/day for panels over 37m in depth. These shallow panels are largely excavated through soft ground using grabs, while the panels over 37m deep were predominantly excavated through Jurong Formation using cutters.

For Prince Edward Road Station, the rock typically becomes less weathered with depth, increasing in strength and density, and resulting in longer excavation times. However, the specialist also reported that the variability of the Jurong Formation at Prince Edward Road increased the overall excavation duration. The cutters use different attachments depending on the rock strength and consistency; as the ground conditions vary between SV, SIV and SIII, time is lost when changing between the different cutting attachments – as shown in Figure 5. If the weathering grade of the rock changes frequently, this process can reduce the excavation productivity.

Figure 5: Changing of cutting tools on reverse circulation rig.

For shallow diaphragm wall panels (20-25m) the productivity increases considerably with panel length– averaging 1.5m/day for 3m panels, rising to 2.9m/day for 6m panels. This may be attributed to the time taken to install and remove stop-ends at the panel joints, an activity that is proportionately more frequent for shorter panel lengths. This trend is less prominent for deeper panels where stop-end installation/removal contributes less to the overall excavation duration.

The average trench excavation rate for the whole of Prince Edward Road Station was 0.8m linear metres per day.

The data suggests a linear relationship between the panel volume – taken as the theoretical excavation volume – and the duration taken to complete the de-sanding process (Figure 6).

For Prince Edward Road Station, the de-sanding rate appears to have been determined by the pump and plant capacity used to process the bentonite slurry. The de-sanding units deployed had a rated capacity of 250m3/hr; however that actual output depends on the density and viscosity of the feeder slurry.

As a planning tool, it may be assumed that the de-sanding duration is directly proportional to volume and is not affected by panel geometry.

The relationship between the cumulative reinforcement weight and installation duration is shown in Figure 7.

As a planning tool, it may be assumed that the installation duration is directly proportional to the total reinforcement weight. For example, 100t of reinforcement is likely to take 10-14 hours to install into the trench.

The duration taken to place concrete within each panel is shown in Figure 8. The blue data series shows the theoretical concrete volume based on the panel geometry, and the orange data series is the actual volume of concrete used; the difference between these two sets of data represents the volume of overbreak with the panel.

The data shows a non-linear relationship between volume and the duration taken to place the concrete, with larger panels achieving a faster casting rate. This is because several tremie pipes can be inserted into longer panels allowing multiple concrete trucks to discharge simultaneously. At Prince Edward Road, a single tremie pipe was used for concrete placement in panels up to 4m in length, and two tremie pipes were used for longer panels; the resulting increase in casting rate for longer panels is reflected in Figure 8.

For irregular shaped panels, such as corners and intersections, it is sometimes necessary to install three or more tremie pipes6; however, this was not required at Prince Edward Road.

Overbreak is not a direct measure of productivity; however, increased overbreak results in additional spoil excavation, more bentonite slurry to be processed and an increase in concrete volume. The following analysis will study the effect of panel thickness, length, depth and excavation duration on overbreak.

To assess the effect of panel thickness, the overbreak volumes were compared for panels of similar length and depth but varying wall thickness (Table 3).

Table 3: Average overbreak volume for panels 45-50m in depth.

The average overbreak volumes were compared for 1.2m and 1.5m thick panels of 45-50m depth (sample size of 144). The average volume of material lost from the trench perimeter was approximately equivalent for 1.2m and 1.5m thick panels; similar behaviour was observed in 0.8m, 1.0m and 1.2m thick panels of 20-25m depth. Therefore, for panels of similar depth and length, the data suggests that panel thickness does not have a significant effect on overbreak volume.

Figure 9 shows the percentage of overbreak relative to panel length for excavation depths of 45-50m and 20-25m (sample size 199). Both data sets demonstrate a reduction in overbreak when the panel lengths were increased from 3m to 6m – typically reducing by 3-5%. One possible explanation is that 3m panels have a greater surface area relative to volume; any overbreak at the panel ends is overcut during excavation of the adjacent panel, therefore as the number of panel joints increases so does the cumulative volume of overbreak.

Both Figure 9 and Table 3 also indicate a relative increase in overbreak for panels 4-5m in length. The standard excavation tools used at Prince Edward Road were 2.8m in length, therefore 3m and 6m panels allow for 1 and 2 full “bites” respectively (a central trimming bite may be required for 6m panels). 4-5m panel lengths are not preferred because the second bite is partial, requiring the excavation tool to cut asymmetrically, as illustrated in Figure 10. This excavation profile is a less desirable because the grab/cutter is not confined and can move more freely within the trench.

Movement of the unconfined excavation tool within the trench may have resulted in the increased overbreak rates observed in 4-5m length panels.

Using the same sample group as Section 4.2, the effect of excavation duration on overbreak volume was studied for panels of similar length and depth. For Prince Edward Road, the data did not support a correlation between excavation duration and the volume of overbreak. Provided that good excavation practice is followed, and trench stability not compromised, excavation duration did not impact overbreak rates.

Figure 11 shows the relationship between excavation depth and overbreak for all diaphragm wall panels (sample size 274). For all panel lengths, a reduction in overbreak percentage was observed with increased panel depth. For example, 3m panels saw overbreak rates reduce from 6-35% at 20-25m depth, to 3-21% at 60-65m panel depth; similar trends were also observed for longer panels. As the rock quality at the site generally becomes more competent with depth, a reduction in overbreak rates can be expected. In addition, overbreak rates may reduce at depth as the head of stabilising fluid increases.

The productivity analysis undertaken in Section 3 showed that the thickness of the diaphragm wall panel did not have an appreciable effect on the excavation rate at Prince Edward Road Station. For shallow panels (20-25m) there was a significant increase in excavation rate as the panel length increased from 3m to 6m – measured in linear metres per day; however, this trend could not be substantiated for panels over 37m in depth.

As anticipated, excavation rates decreased with increasing panel depth, with a significant drop in productivity observed in panels over 37m deep. This represents slower progress when excavating through rock, which is further compounded if the weathering grade is highly variable and requires the cutting tools to be changed frequently.

Analysis of de-sanding duration shows a linear relationship with panel volume, suggesting that productivity was determined by the pump and plant capacity. Similarly, the duration of reinforcement installation was directly proportional to cumulative cage weight.

The data shows that the rate of concrete placement increases with panel volume because longer panels allow several tremie pipes to be used simultaneously. Figure 8 also illustrates the increase in concrete volume caused by overbreak, leading to additional cost and longer casting times.

For Prince Edward Road, the volume of overbreak was not affected by panel thickness or excavation duration. However, increasing the panel length from 3m to 6m could reduce the percentage of overbreak by 3-5% - thus reducing soil disposal and concrete material costs. These benefits are only applicable if the trench stability can be safely maintained for a longer panel and may not be appropriate if the diaphragm wall is located near to sensitive structures. Panels 4-5m in length, requiring a partial bite of the excavation tool, appear to provide the least favourable overbreak conditions.

Overbreak rates were seen to reduce with increasing panel depths. This is consistent with a general improvement in rock competency and greater head of stabilising fluid within the trench.

Table 4 provides an example of how the productivity information contained within this paper could be used as a planning tool for similar diaphragm wall works.

The reader should note that the findings are derived from a single project with specific ground conditions, and they may not be directly applicable to other projects with different geology.

1. Land Transport Authority, 2017. Geotechnical Interpretative Baseline Report – Contract 885 – Station CC32 and bored tunnel from Station CC31 to Station CC32.

2. German Institute for Standardisation. DIN 4126:2013 Stability Analysis of Diaphragm Walls.

3. Land Transport Authority, 2010. Materials & Workmanship Specification for Civil and Structural Works, Chapter 6 (available at www.lta.gov.sg).

4. 6. Institution of Civil Engineers, 2007. ICE specification for piling and embedded retaining walls – second edition, pg 180, 182-184.

5. Malcolm Puller, 2003. Deep Excavations: a practical manual – second edition, pg 121, 128-130, BSI Standard Publication, 2015. BS EN 1538:210+A1:2015 Execution of special geotechnical work – Diaphragm walls.

Defense Science and Technology Agency, 2009. Geology of Singapore, 2nd Edition.

Want to read more? Subscribe to GE’s enewsletters and follow us on Twitter and LinkedIn