Feedback from readers appreciate the article and there is a contribution with additional details about the collapse and rehabilitation of the Estì headrace tunnel in Panama and a response to the recommendation for independent checking for hydropower waterway designs and the use of different methods for designing watercourses subject to the internal pressure and velocity of water through the generation plant and through diversion tunnels around dam construction sites.
See the full Feedback on the Feedback page and at the bottom of the article.
Add to the discussion with your comments via the Feedback form.
Design and construction of hydropower tunnels is based on significantly different design criteria and practices in comparison to other tunnels for civil engineering infrastructure. Given that they are subjected to internal and dynamic loading during operations, the performance of these tunnels can only be monitored regularly by pressure instrumentation. Dewatering of the watercourses for physical inspection require operation outages and the loss of generation and associated revenue. Dewatering may also cause a negative impact to the integrity of the tunnel.
Design of a hydropower tunnel should therefore be based on a thorough review of the possible worst case operating conditions. Reservoir schemes typically maintain constant pressures in watercourses with run-of-river schemes subjected to seasonal fluctuations and their associated significant pressure variations. Projects that are designed on multiple intakes within a large watershed can be expected to result in significant pressure oscillations within a watercourse.
For predominantly unlined hydropower tunnels, the final design is only completed after a thorough evaluation of the encountered geological conditions and the design and application of any final lining needs at as required and at relevant sections.
Since 1995, more than 300km of low-pressure hydropower tunnels have been excavated using double shield TBMs in conjunction with precast concrete segmental linings. These have been completed successfully with no problems reported to date.(1) Where there exists the risk of squeezing rock conditions it may be more appropriate to adopt a single shield TBM. These approaches have achieved good rates of production as well as providing greater safety against risks associated with variable and difficult geological conditions including fault zones and rockbursts due to release of stresses in headings beneath high overburdens.
It is also important for clients to share all relevant information about the planned modes of system operations, especially for design-build delivery projects, so that all modes of operation can be fully understood and considered and incorporated into the final design of tunnel linings.
Table 1. Hydropower waterway collapses in the past decade | |||||
Project | Year | Diameter | Length | Operations period prior to collapse |
Repairs |
Glendoe | 2009 | 5m | 8km | 8 months | Bypass |
Rio Esti | 2010 | 10m | 5km | 9 years | Concrete lining |
La Higuera | 2011 | 6m | 16km | 9 months | Bypass |
Shuakhevi | 2017 | 6m | 18km | 3 months | Injection and additional support |
Ituango | 2018 | 16m | 2km | During construction | Unknown |
The internal static and dynamic internal operating pressures are important aspects to consider for the design of pressure tunnels. Hydraulic transient analyses should be performed by the hydraulic designers to confirm the magnitude and variation of internal operating pressures to be expected during normal and non-normal operating periods.
Notable historical collapses of hydropower watercourses that occurred in the 1950s and 1960s include the Kemano project in Canada, the Snowy Mountain Project in Australia and at the Lemonthyme project in Tasmania, Australia. The collapse at Kemano occurred after three years in operations and included a large 25,000m3 collapse. A total of three collapses after four years of operations occurred in the Snowy Mountains scheme and included a total volume of about 3,000m3. The collapse at the Lemonthyme tunnel occurred after five months of operations and involved a volume of about 1,500m3.
In the last decade, a set of five notable hydropower tunnel collapses occurred in Scotland, Panama, Chile, Georgia and Colombia (Table 1).
Along with insurance investigations, legal proceedings resulted for most of the tunnel collapses due to the substantial repair costs involved and the loss of power generation revenue due to the shutdowns. Most of these collapses resulted in financial costs exceeding US$250 million and a loss of generation for up to 24 months.
The 100MW Glendoe installation is located on the southwest side of Loch Ness and includes an 8km x 5m power tunnel excavated on an upward gradient of 12% using an open gripper TBM. The internal operating head pressure is 600m and the flow capacity is about 18.6m3/s.
The project was constructed under a design-build contract with a specified design life of 75 years. The hydropower plant was to be operated intermittently as a stop-restart system subject to reservoir water supply availability. This would cause highly variable internal pressures to the tunnel.
The 120MW Rio Esti hydropower project is located in Chiriqui Province of western Panama, about 30km north of the city of David, at an elevation of about 220m. The scheme comprises the Chiriqui and Barrigon Dams and reservoirs, a 6.5km canal, a 4.8km x 10m diameter headrace, and a surface powerhouse. The total head is 112m with a flow capacity of 118m3/s.
The project was procured under a compressed design-build contract of 33 months to meet a power generation purchase agreement deadline. The headrace was designed and constructed as a near-fully shotcrete lined excavation since there was inadequate time in the schedule for a final concrete lining, even from multiple access adits. The geology on the tunnel comprises horizontally bedded volcanic sedimentary rock including mudstones, andesites, tuffs, and agglomerates, with a depressed groundwater table below the alignment. The vertical alignment was varied as much as possible to avoid intersection of the tuffs. A major collapse near the surge shaft and with an estimated volume of 14,000m3 caused near total blockage of the tunnel.
The failure of the Rio Esti headrace is believed to be due to the first time saturation of the weak volcanic sedimentary rocks that increased the loading conditions on the shotcrete lining at various locations. The acceptable operation of the hydro station for nearly seven years is remarkable given this design. The repair of the tunnel comprised re-profiling and the installation of steel ribs and casting a final concrete lining for the entire 4.8km length. The total outage for the repairs was about 23 months.
The 155MW La Higuera hydropower project is a run-of-river scheme located within the Tinquirirca Valley about 150km south of Santiago. The 16km headrace forms part of the downstream component of a three-project cascade scheme. The tunnel was constructed under an engineering, procurement, construction (EPC) contract using drill+blast from multiple access adits, at a relatively low gradient and with a shotcrete final lining throughout most of the tunnel.
A prominent geological fault was intersected at the downstream section of the tunnel where nominal and final support comprised rock bolts, mesh, and shotcrete. An inferred weak geological formation or layer can be identified on the surface near the location of the collapse. The headrace operated for about nine months prior to collapse.
The weathered rock, associated with the geological fault, created a volume of about 12,000m3 at the collapse location⋅(2) Tunnel repairs comprised the construction of a 240m long bypass tunnel. The total outage for tunnel repairs was about 21 months.
The 181MW Shuakhevi hydro project is a run-of-the-river plant located in southwest Georgia and includes waterways of more than 35km, two dams and two surface powerhouses. Drill+blast excavation of the tunnels started in 2013 from multiple intermediate access adits.
Georgia has complex and disturbed geology due to its location between the Eurasian and Africa-Arabian tectonic plates. The rock types noted along the tunnels comprised highly disturbed basaltic andesites, conglomerates and volcanoclastic sandstones as part of a major geological syncline oriented sub-parallel to the downstream headrace.
Operation of the plant started in mid-2017 and after about three months a series of collapses and other damage occurred in the main 19km downstream waterway. Most of the collapses were reported to have fully blocked the tunnel.
Repairs originally planned a bypass tunnel at one of the major collapse areas but were eventually based on grout injections and additional rock support at the collapse areas. The total outage for the repairs was about 30 months.
The 2400MW Ituango hydro project, on the Cauca River about 170km northwest of the city of Medellin, comprises an underground powerhouse with eight intake tunnels and shafts, four tailrace tunnels, and two diversion tunnels to assist construction of a 225m high earth core rockfill dam. An additional third diversion tunnel was introduced during construction to improve the overall construction schedule.
In May 2018 during the rainy season a collapse sinkhole of 14m across occurred near the portal of the third diversion tunnel as the only operating diversion after the two main diversions had been plugged as part of the impounding of the reservoir. The incomplete dam spillway, together with the rapidly increasing reservoir level behind the dam resulted in an emergency decision to release a large volume of water through the intake tunnels and through the nearly completed underground powerhouse, resulting in catastrophic damage.
The diversion tunnel collapse was of an unusual circular or crater shape of inadequate competent rock above an underground excavation. The location of the third diversion portal area was within a topographic depression upstream of the two main diversion tunnels. Based on photographs, the rock cover above the collapse area was believed to be less than 10m, which is less than the width of the tunnel. The total volume of the collapse was estimated to be 120,000m3.
Collapse of the third diversion tunnel is considered to be a result of the high 50m head of internal hydraulic pressure associated with the high water levels of the river causing saturation of the surrounding overburden and imposing additional load on the tunnel support. The third diversion was designed as a shotcrete lined tunnel with conventional rock support without any concrete lining, as would have been installed as common practice for these important elements. The third diversion tunnel was not designed to accept the additional loading conditions associated with high water levels that occurred during the rainy season.
The collapse of the diversion tunnel caused significant consequential damage to the underground works that resulted in the largest insurable construction industry incident of US$ 1.8 billion with repairs now ongoing to the underground works with diversion occurring through the completed spillway of the dam.
Upon review of the details, it is apparent that the main causes of collapses have been due typically to design errors - not construction defects – due to the incomplete identification of weak geological zones including medium to large faults as well as minor fracture zones containing dissolvable minerals including clays, anhydrite and gypsum. The technical disclosure of these important conditions by the geological team to the project design team is vital to avoid misunderstanding of, or non-recognition of the cyclic loading conditions of oscillating internal pressures, particularly for peaking operations, and the need for high capacity tunnel lining designs to effectively protect and stabilise these conditions for future hydraulic system operations.
The hydropower industry has often assigned the responsibility and relied upon the judgement of site geologists to make the final lining design decisions including the simplified application of rock mass classifications without fully appreciating and understanding the nature of future hydraulic operations and imposed loadings for the tunnel. In addition, as per other significant and critical engineering structures, there appears to be no independent checking of the final lining designs.
The most common and safe method for the repair of a tunnel collapse of limited extent is with the construction of a bypass tunnel and plugging off the original tunnel between the bypass as completed at Glendoe and La Higuera.
The use of shotcrete for long-term stability and support of major geological faults within hydropower waterways is not considered to be adequate, particularly for peaking power plants and where highly variable internal pressure oscillations can cause preferential scour, saturation, and deterioration of such final linings. New tunnel support design standards comprising the use of concrete linings are considered necessary for the long-term stability and support of major geological faults in hydropower watercourses.
In some regions where geological conditions are highly disturbed or non-durable, it may be warranted to provide concrete linings for a significant length or the entire tunnel.
The following conclusions are considered to be applicable to the occurrence of the recent hydropower tunnel failures:
Hydropower waterway designs and recommendations
Feedback from: Fabrizio Bove, Project Manager, Seli Overseas
Thank you for the interesting analysis carried out and good suggestions shared.
Since I had the challenging opportunity to work for the rehabilitation of the Estì headrace tunnel in Panama, I would like to add and share some issues for further considerations:
To: Failures due to the incomplete identification of weak or critical geological zones
In the case of Estì, the failure occurred mainly because horizontally bedded weak layers that ran parallel to the tunnel axis in several areas and were not visible during excavation. In other words, during the face inspections the rock mass had been recorded as self-stable and sound with high value rock mass indexes, thus leading to a choice for a softer lining of shotcrete. As a result, only a complete overview, focusing of the stabilization of the on-going advance core and fully appreciating the in-situ 3D geology and the nature and pace of future hydraulic loading for the tunnel, could improve the choice of the permanent lining and stability.
To: A general industrialized approach to excavation for example, using TBMs with standard reinforced precast segments or conventional methods with cast-in-situ
The suggestion is that these should have led to better results, avoiding the uncertainty of shotcrete for long-term stability. Most of the time the application of quality shotcrete relies on operators and fibres randomly distributed. More difficult quality control measures are also required especially for larger diameter tunnels.
To: An “ad hoc” approach and particular care may improve the results in case of critical areas
Where the biggest collapse of about 14,000m3 occurred in Estì, at least three factors could have hampered the situation:
To: Choice of contract type
An EPC contract was used for the main project procurement and also for the Estì rehabilitation phase in 2011 and 2012 and with good achievements. Obviously however, a stronger control by the Engineer is needed in these cases.
To: A contractual “risk-sharing approach” between Employer and Contractor
In the case of unforeseen and unforeseeable events this can be the best option, both in terms of cost control and timely reaction and application of solutions.
Best regards,
Fabrizio Bove
Project Manager
Seli Overseas
Feedback from: Chris Breeds
Great article Dean. Very informative and great examples.
Best regards,
Chris Breeds
Feedback from: Nick Barton, Independent Consultant
Dear TunnelTalk,
It is always interesting to read of tunnel failures, from which we all learn. I have two points in response.
In his interesting and well-illustrated review, Brox has suggested using “other than rock mass classification methods” for application of final support and linings. There is nothing wrong with this suggestion except that it ignores the contribution of such methods to thousands of kilometres of hydropower tunnels and considerable cost savings for owners. If mistakes are made in application of the methods due to oversight, especially in the case of faulted rock, then lessons need to be learned by those involved. Incidentally the Ituango case, the failure erosion cone is much larger than described and is a special case of optimism in diverting water with a peak velocity of up to 10m/sec around a remarkably tight bend, both of which are entirely different to the typical 1.5-2.5 m/sec velocities in the case of hydropower tunnels. Design for velocities of 1.5-2.5 m/sec have been and are the basis of Q-system case records. With Eda Quadros, I have prepared a paper for Eurock 2020 titled Some lessons from single-shell Q-supported headrace and pressure tunnels that may not now be presented due to Covid-19 postponements but may become available in published proceedings of Eurock 2020 planned for 15-19 June 2020 in Trondheim, Norway.
Secondly, Brox recommends independent checking of waterway tunnel designs. I agree that this could, in principle, be valuable. I have reservations however based on what is available outside of the use of more careful rock mass characterization and use of empirical methods, like the Q-system which probably has the most relevant database. It should not be forgotten that there are thousands of kilometres of such waterways, and many hundreds (actually thousands) of economic projects as a result of the single-shell type of support, which, as pointed out by many, needs to consider the intended use of the tunnel. As indicated above, if a water velocity, as in the case of a river diversion, is chosen by a designer that is well outside the database (for example 10m/sec as compared with a conventional 2m/sec velocity), then one is asking for potential trouble, if the tunnel support, also of the invert, is not dimensioned accordingly.
Having reviewed many projects over the years, and having experience of a specific international court case, it is evident that numerical modelling is frequently too much relied upon. These experiences have also demonstrated the severe limitations of popular numerical modelling methods, those that produce the colourful appendices (and maybe pay-raises), might well be used by engineers engaged to carry out independent checking with the assumption that they are more reliable and sophisticated than rock mass classification and empirical methods. The exaggerated so-called plastic zones of supposedly jointed models, or of the simpler continuum models, has led to jokes about needing snow shoes so as not to sink in the models as they show quite different behaviour to more reliable models with longer track records.
Of course, it is also possible to misuse the more reliable codes and their longer track records by, for example, modelling with exaggerated joint continuity. There are hundreds of examples of this. In fact, it is found that a super-simple empirical model, linking deformation, tunnel span and the Q-value, provides a far more accurate check of potential deformation, as monitored subsequently at the completed project, than the exaggerated models. With more realistic engineering- geologist-generated joint continuity and its digitization, a more realistic and smaller deformation is predicted which agrees well with the simple empirical check.
Best regards,
Nick Barton
Independent consultant
Norway
|
|