Looking at tunnel roughness

3 April 2008



Research in Kárahnjúkar headrace is providing fresh insight into tunnel wall roughness to estimate headloss. Report by Patrick Reynolds


The TBM drives for the headrace tunnel of the Kárahnjúkar hydropower scheme, in Iceland, found highly varied geological conditions, particularly with the large water inflows that were suffered in some parts of the bores. Beyond that tough construction experience, however, different lessons are now being learned that should be useful for tunnel lining works in a wider range of water projects thanks to client-sponsored research into surface, and hence hydraulic, roughness.

The client, national power company Landsvirkjun, was naturally focused on headloss in the hydraulic system not only for fundamental economic reasons but also due to the potential scale of lost energy, given that the 7.2m-7.6m diameter headrace tunnel is 39.6km long. Prior to the research, design estimates used in planning the project estimated the hydraulic friction losses at anywhere between approximately 60m-95m, which is about 10%-17% of the nominal gross head of 600m under flow rate of 144m3/s, or full load conditions.

Every percentage point saved in headloss, and even the shavings of such, represents energy for sale to a power-hungry aluminium smelter operating in a hot commodity market, and also the electricity grid. Landsvirkjun aims to generate average energy 4600 GWh/year. Beside the economic benefits to the client, though, there would also be gains by using the greater knowledge of the insitu tunnel roughness to fine-tune operational rules for the hydropower system.

Beyond the immediate client, other project sponsors with water tunnels in which headloss is of key importance can also draw upon the research to help secure additional economic and operational benefits. To gain such, in the most effective way, they will require the supply side services of designers and contractors to absorb the research findings to help estimate, monitor and adjust the surface roughness of tunnel lining works during the construction phase, especially so if data were to indicate that consequent energy losses would be outwith design estimates. The data could also inform budgetary options and construction phase logistics should opportunities arise, or be pursued, for improving wall roughness within the estimated range of headloss.

Research project

The 690MW Kárahnjúkar hydropower project marked the first use of TBM tunnelling in Iceland, and so there was not a large pool of domestic experience to draw upon to execute the works.

Design and construction of the scheme has been a significant international effort, drawing upon services from local and foreign consultants in the client’s design engineer Kárahnjúkar Engineering JV (KEJV) – VST Consulting Engineers, Pöyry, MWH, Rafteikning and Almenna Consulting Engineers. The design programme commenced in 2000. As local practice prevents the designer from supervisory duties, a consortium was hired as owner’s representative - VIJV, which comprises Mott MacDonald, Linuhonnun, Hnit, Fjarhitun, Sweco, Norconsult and coyne-et-bellier.

The main contractor undertaking the KAR-14 headrace and Jökulsá diversion tunnel works under a re-measurement contract, and other works, is Impregilo. The headrace and diversion tunnel have been excavated through varied volcanic and sedimentary rocks mostly by TBM with some drill and blast, though slightly more than initially planned. Three TBMs – two 7.2m diameter main beam shields and a 7.6m diameter machine - were supplied by Robbins for the project. Excavation of the headrace commenced in mid-2004 and finished late 2006, and the Jökulsá bore is almost complete.

In total, the US$1.3B scheme will have approximately 73km of hard rock tunnel, mostly unlined but with stretches of shotcrete walls, and an underground powerhouse.

The view was that the available technical data and literature on roughness of long TBM-bored water tunnels was incomplete and of limited reliability in spite of past experience of design and operation of such bores internationally. In particular, there was a lack of good information on unlined TBM-driven tunnels especially through volcanic rock.

The idea for the tunnel roughness measurement programme came from the hydraulic coordinator, Gunnar Gudni Tomasson of VST, and lead headrace tunnel design engineer, Joe Kaelin of Pöyry Energy. They also supervised the studies, undertaken by others in KEJV - local firm VST in conjunction with Pöyry – and completed during the headrace excavation. The research effort was supported by VIJV and Impregilo.

A different, established approach - Rønn’s “IBA” method, from Norway - was used to assessing headloss in unlined tunnels excavated by drill and blast. The method is one of many approaches (eg. Rahm, Colebrook, Huval, Priha, Reinius, Wright, Johansen, Solvik and Czarnota) and it is based on measurements of cross section and longitudinal geometry over 20m-25m long portions of tunnel to calculate wall roughness and, hence, equivalent hydraulic friction.

To prepare for the studies on the TBM drives, the team drew upon the methodology of wall roughness and hydraulic headloss studies, established by Pegram and Pennington in the report to the Water Research Commission by the University of Natal. It included the case study of Delivery Tunnel South of the Lesotho Highlands Water Project (LHWP), but looked at different surface types (sandstone, granite, shotcrete and concrete) in TBM-driven tunnels in general. Laser measurement was used in four tunnels, two in LHWP.

Measurements – visual inspection

At Kárahnjúkar, the aim was to collect physical measurement data during construction along the entire length of the headrace. Visual inspections of tunnel walls were conducted at 50m intervals, exactly, over 2005-6.

During visual inspections, a 1m-wide strip of rock was inspected below springline, at springline level and in the crown at the side of the tunnel, opposite the mucking out conveyor, to categorise intact rock or shotcrete into different roughness classes – smooth, medium and rough. The classes were defined to represent the range of observations and were based on measurements of the maximum average deviation of the wall surface from a straight bar ruler - 40cm long for rock surfaces, 80cm long for shotcreted surfaces; they are, in principle, independent of geology. Additionally, the surface below the springline level was photographed.

Specifically, the smooth, medium and rough categorisation for rock surfaces resulted from maximum average deviations from the straight bar ruler of <5mm, 5mm-14mm and 15mm-24mm, respectively. The corresponding thresholds for shotcrete lining were <10mm, 10mm-25mm and >25mm, respectively.

It should be noted, though, the processing of the data from the observations took into account planned finishing work subsequent to inspections, eg further shotcrete applications or treatment and cleaning of surfaces. The researchers allowed for these estimates based on size of the sections involved and observed workmanship and quality control on executed works.

Observed large-scale rock features, such as joints and/or pockets, were classified into four categories by the depth and number. In the lowest category, representing rock with no such features, the classifications of smooth, medium and rough fell within the limits of the deviation measurements from the bar ruler, as noted above. The measurement bands for smooth, medium and rough in each of the three other, rising, categories are up 40mm, up to 100mm and in excess of 100mm, respectively.

To aid consistency, the same hydraulic engineer was responsible for classifying all the surveyed surfaces. In total, almost 90% of the TBM drives were inspected visually, equating to 1893 observations at 631 locations. The remaining 3.9 km of headrace comprised discontinuous, inaccessible stretches during the inspection period.

In summary, a quarter of the inspected tunnel was shotcrete lined, and those sections classed as: smooth (14%); medium (64%); and, rough (22%). While rebound shotcrete prevented about 4% of the unlined rock surfaces from being classified, those assessed were classed as: smooth (26%); medium (65%); and, rough (9%).

In terms of the geological strata, those rocks with proportionally more rough surfaces ranged from pillow lava, tillite, and cube jointed basalt in decreasing order to andesite, scoria, pillow breccia, scoracious basalt, conglomerate, olivine basalt, porphyric basalt, tholeiitic basalt, and sandstone/conglomerate, and sandstone.

Finally, at the other end of the range, siltstone and sandstone/tuff were classed as having no rough surfaces. However, it should be noted there were few observations of andesite, siltstone and tillite.

Observations of the range of joints and large scale irregularities concluded that about 45% were sparse and/or shallow, a quarter were dense and/or deep, a quarter were free of joints, and approximately 5% of the surfaces contain large rock break-out.

In terms of geological strata, rock with many and/or deep joints were found, in decreasing order of importance, associated with: cube jointed basalt, pillow lava, tholeitic basalt, olivine basalt, tillite, andesite, porphyric basalt, scoria, pillow breccia, scoracious basalt, sandstone/tuff, conglomerate, sandstone/conglomerate, sandstone and siltstone.

Joint-free surfaces were found in association with: siltstone, sandstone/conglomerate, sandstone, conglomerate, sandstone/tuff, scoracious basalt, scoria, tillite, porphyric basalt, cube jointed basalt, olivine basalt, tholeitic basalt, pillow lava and andesite.

Measurements – laser scan

Laser scans were executed at approximately 10% of the visually inspected surfaces. Rock surfaces were measured - using a FARO Scan Arm portable coordinate measuring machine - below springline level, and shotcreted surfaces above springline level as the upper parts were lined when measurements were taken.

About 600 profiles were scanned at 73 different locations over nine months. The sites to be scanned were chosen from photographs to represent the different roughness classes, and were dry, or had to be dried, before the scan. The scanned surfaces represented nine roughness categories - four to reflect the range within each of the classification groups of smooth, medium and rough. In the same approach, the shotcrete walls were represented by three roughness categories.

The support platform for the scanner was bolted to the wall and the head of the scanner fixed to a 1m long bar. Steady motion scanning was facilitated by a manually-operated cogwheel.

Scans were made of the walls in 40mm wide strips, each strip comprising 640 points. The coordinates are scanned with an accuracy of 0.1mm and point spacing of less than 0.25mm. At each survey site, the scanner measured two 40mm wide by 1m long strips, separated vertically by about 100mm. Four profiles with a vertical separation of 8mm were extracted from each scanned strip, which provided eight roughness profiles at each scan site. The average sample spacing was 0.1mm.

Data analysis

The data obtained from the and laser scan measurements were processed according to the method developed by Pegram and Pennington, which transformed the data from physical to hydraulic roughness. These drew upon three parameters: equivalent sand grain roughness, ks, also referred to as Nikuradse’s equivalent sand grain roughness; a dimensionless friction factor, f, associated with the use of the Darcy-Weisbach flow resistance formula; and, a flow resistance coefficient, associated with the alternative use of the technical Manning (or Strickler) flow resistance formulae. The relationship between f and ks is expressed by the Colebrook-White equation.

Wall roughness in TBM-bored tunnels has a wave-like structure that varies with geological strata as well as the operation of the cutterhead, though data on TBM advance rates were not explicitly compared to intact roughness. The researchers note further that the scale of the roughness arising from cutting speed or method is usually an order of magnitude smaller than that due to the geology. They add that the regular waves that were ground into strata were only visible on smooth rock surfaces.

Still, the wave properties can be calculated from the laser measurements, which contribute to the headloss calculation by establishing the dominant wave length and taking its ‘bump’, or wave height, as an approximation to the equivalent sand grain roughness.

In the initial design estimates, the equivalent sand grain roughness for unlined TBM drives was taken as 10mm, and 20mm for shotcrete walls. The interpreted measurements data indicated that 40% of the unlined headrace had a larger roughness value and 60% lower. The average value for the shotcrete walls was determined to be lower than estimated, at 17mm.

It should be noted that the headloss effects resulting from directional changes in the tunnel, such as at bends, are calculated separately as singular energy losses as commonly established in hydraulic system analyses. Further, in assessing the possible effects of variations in tunnel cross-section, this is considered to be negligible for TBM-driven bores while those in drill and blast sections are accounted for in established analyses, such as the IBA method.

Frictional headlosses corresponding to the full rate, design discharge were calculated for each roughness category using the friction factors given by the Colebrook-White formula and the Darcy-Weisbach equation. Based on the calculated roughness data, and taking the 7.6m diameter TBM to have bored 14.7km and the two 7.2m diameter machines to have excavated a total of 20.8km, the specific headloss per unit length of the tunnel was determined. The researchers note that the calculations showed specific headlosses for shotcrete sections to be similar to those for rock with large break-out, partly due to the minor throttling effect of its thickness slightly reducing the internal diameter of the tunnel.

Based on the measurements and data analyses, the research with its elected methodology determined that the overall headloss from friction in the TBM-bored section of the headrace was 64m with a tolerance of 10%, which puts it at the lower end of the initial design estimate. The corresponding initial estimate, based on average values of roughness coefficients, is 71m.

Verification work is now underway to derive the actual headloss in the headrace during operation of the power plant, which started-up last year. While the headrace is not yet carrying the full, design flow rate, only 100m3/s, so far the direct measurements show that operational headloss is 6% less than that predicted from the roughness measurements. As the discharge rate increases the headloss measurement will become more accurate, the researchers said.

R&D and applicability

The researchers say the key area for further research is to improve the accuracy of matching the scanned data of wall roughness to the actual equivalent sand grain roughness, as this is independent of tunnel diameter. While the operational headloss information at Kárahnjúkar verifies understanding of the relationship it is, at the same time, a blunt instrument, to a degree, though the best yet; it is comparing the actual and estimated values at the ‘macro’ system level and does not, for practical and other reasons allow for operational investigation in individual tunnel sections.

Their hope, therefore, is to gain backing and support to undertake a programme of laboratory tests to further refine the knowledge of how equivalent sand grain roughness relates to actual wall surfaces. Such research will benefit design and construction of all water conveyance tunnels, no matter the specific combination of geology, add the researchers.

In the meantime, the research as it stands also has practical application for all water tunnels, particularly so for those in which the economics of a project depend to a high degree on estimates of headloss, or the energy loss is an important design factor for other reasons, such as calculating transients, say the researchers.

They add, though, that while the research data from the studies only apply directly to the geology at Kárahnjúkar and would have value to other drives traversing volcanic geology, the conclusions established with regard to the roughness of different rock formations and shotcrete surfaces are applicable to similar strata and tunnel lining elsewhere.

The research findings, therefore, can help confirm design assumptions through the construction phase of a project. The findings can also assist in managing any changes to tunnel lining works that might possibly be required or could arguably be of longer-term benefit to a client.

There is always the additional possibility – especially as research advances - of following up the construction phase to measure actual headloss not only to take the opportunity for a further, general check of the headloss prediction system but also, possibly, to ensure the contracted performance has been delivered.

IWP&DC would like to thank the research team for the briefing, especially Kristin Martha Hakonardottir of VST, Gunnar Gudni Tomasson of Reykjavik University and VST, and also their co-authors in a recent research paper on the work, Bela Petry of Delft Netherlands and Bjorn Stefansson of Landsvirkjun. They also acknowledge the initiative for and supervision of the research by Joe Kaelin of Pöyry, the visual inspections by Snorri Gislason, a geologist at VST, and the laser scan work by René Fretz, a survey engineer with Pöyry.


The laser scanned vertically-separated strips of wall surface - 1 The laser scanned vertically-separated strips of wall surface - 1
Equivalent sand grain roughness calculated against rock type Equivalent sand grain roughness calculated against rock type
Specific headloss ranges in different diameters of the headrace Specific headloss ranges in different diameters of the headrace
Visual inspection Visual inspection
Figure 1 Figure 1
The laser scanned vertically-separated strips of wall surface - 2 The laser scanned vertically-separated strips of wall surface - 2
Caclucalted equivalent sand grain roughness from survey data Caclucalted equivalent sand grain roughness from survey data
Figure 5 Figure 5
Figure 6 Figure 6
Figure 3 Figure 3
Survey findings Survey findings
Figure 2 Figure 2
Figure 4 Figure 4
Following the visual inspection - 2 Following the visual inspection - 2
Proportions of roughness class measurements Proportions of roughness class measurements
Following the visual inspection - 1 Following the visual inspection - 1
Karahnjukar  location map Karahnjukar location map


Privacy Policy
We have updated our privacy policy. In the latest update it explains what cookies are and how we use them on our site. To learn more about cookies and their benefits, please view our privacy policy. Please be aware that parts of this site will not function correctly if you disable cookies. By continuing to use this site, you consent to our use of cookies in accordance with our privacy policy unless you have disabled them.