Learning lessons on RCC design10 September 1998
Uncontrolled seepage and cracking, which were once commonplace in RCC projects, are now the exception, according to Randall Bass*
Since 1982, 29 roller compacted concrete (RCC) dams 50ft or more in height have been built in the US, and there are two others under construction. Lessons learned from these projects over the last 16 years have led to significant changes in the design and in the means and methods of construction, particularly in the area of seepage control and collection and crack control.
The earliest dams were generally built as a monolithic structure with no joints and few, if any, crack inducers. Additionally, lift quality was frequently compromised because equipment had to be hauled on and off the lift surface. Both of these factors have contributed to uncontrolled seepage.
The instances of uncontrolled cracking and seepage have not been a safety concern, but they have created an aesthetic and maintenance problem at several of the projects. Repairs have generally consisted of sealing the individual large cracks, and grouting the mass of the dam by drilling through the crest in an attempt to seal cracks and any voids along lift lines.
Today, designers of RCC dams place greater emphasis on evaluating the potential for cracking due to thermal and foundation considerations. They are minimising uncontrolled seepage by providing upstream facing systems designed to be near-watertight and/or providing means to collect the seepage before it exits at the downstream slope. The performance level of RCC dams has improved from the early days because the design community has been willing to exchange information on how their projects have performed and have developed the expertise to address past performance deficiencies.
In the early years, seepage through those RCC dams that maintain a normal pool was generally controlled by using a conventional concrete facing, placed concurrently with the RCC. This first generation of facing systems was 1-3ft thick, unreinforced, and had no joints or waterstops.
Two of the larger projects, Upper Stillwater and Elk Creek, utilised a slipform system which placed both the upstream and downstream facing before the RCC was placed. However, they took different approaches to controlling seepage.
Upper Stillwater used a high flyash mix with a total cementious content of 425lb/ft3 to produce a high-strength very workable RCC mix. The high flyash content (290lb/ft3) minimised the heat of hydration and contraction joints, to accommodate thermal strains, could be eliminated. The Elk Creek design drew from experience at Willow Creek. Waterstop contraction joints were used in the facing and dam: this required the RCC to be placed during cool weather and a bedding mix to be used along the entire surface of each lift. Two foot high lifts were used to speed production and reduce the number of lift joints.
Upper Stillwater leaks through some cracks but there is no significant seepage through the RCC mass. The seepage control measures implemented on Elk Creek were never put to the test because the project was halted at only one third of its height and does not retain much of a normal pool.
The latest generation of conventional concrete-facing designs include waterstop contraction joints, not only at the location of the crack inducers in the mass sections of the RCC but also at 20-30ft intervals along the length of the facing element. The joints aligned with the crack inducers may have a double row of waterstops, with a drain located behind the waterstop to collect any water that bypasses it.
Some designers have introduced a series of face drain holes that are drilled diagonally from the top of the completed dam, intersecting the drainage galley, to collect any seepage through the dam’s mass or through any uncontrolled cracking. The drain holes are angled at 45%, and alternate direction on 10ft centres across the crest of the dam. The New Elmer Thomas dam has such a system, and reports negligible seepage.
The one early exception to using a conventional concrete facing was Winchester dam in Kentucky in 1984. Here a geomembrane attached to precast panels was developed as an upstream facing. This project used a 65mil PVC geomembrane attached to a 4x16ft concrete panel as the primary element to prevent seepage. No measurable seepage through the dam has ever been recorded.
The effectiveness of this early application of a geomembrane led designers to design this type of system in six out of the last eleven projects: all have used PVC, except for one case where LDPE was used. The membrane is generally 65-80mil thick and is attached to a 4in thick precast concrete panel during fabrication. The attachment has been, either by the use of T ribs, moulded with the geomembrane or by a geotextile heat-bonded to one side of the geomembrane. With the T ribs, the geomembrane is placed first at the bottom of the casing form with the ribs facing up into the fresh concrete. When using the geotextile the concrete panel is cast first; while the concrete is fresh, the geomembrane is placed with the geotextile surface against the concrete. The concrete mortar impregnates the geotextile and holds the geomembrane to the panel. Tests have shown that the geotextile application allows the geomembrane to elongate across the entire width of the panel, where the T rib application can only elongate between the ribs, which are less than three inches apart. If appreciable movement is expected, the geotextile attachment approach can tolerate greater movements without compromising the integrity of the geomembrane.
To complete install-ation of the geo-membrane facing system, individual panels of liner are heat-welded together as the precast panels are erected. Welding the geomembrane panels is paramount to the success of the system. Experience has shown that the productivity of the welders is directly proportional not only to cold weather but also the cleanness of the geomembrane. Contaminants such as dirt, grease, form oil, and prolonged UV exposure will slow down welding.
The two most recent dams have used larger precast panel sizes than the 4x16ft previously employed. The larger panels measure up to 6x16ft, which reduces the liner feet of joints in the geomembrane system by 33% and thus the potential for leakage through the welded connections.
The degree of backup elements to the geomembrane facing system vary between the designers: New Penn Forest and C E Seagrist elastomeric sealant was used to seal the horizontal and vertical joints created by stacking the precast panels to form the upstream face; Big Haynes and Buckhorn had a seepage collection system placed between the geomembrane liner and the RCC where a heavy geotextile was placed against the liner with collection piping to direct any collected seepage to the outlet pipe.
The performance of the geomembrane facing systems have been excellent. Seepage rates through those dams with this system have been little to immeasurable, except for Spring Hollow where the lower third of the downstream slope is very wet, with flowing water in many places. Here analysis indicates that most of the water originates in the abutments; it accesses the lift lines at that location and exits at the downstream slope. This is one of the dams that received post-construction grouting after the initial filling to reduce the seepage flow rates. Over 1000ft3 of grout was injected to fill voids in the dam but the seepage was not reduced substantially.
In almost every case seepage flow rates through RCC dams have decreased with time. Siltation and calcification are the major natural reasons for seepage reduction. Calcium carbonate is formed when calcium hydroxide released from the cement hydration process is carried by seepage to a surface in contact with air. The calcium hydroxide combines with the carbon dioxide in the air to form calcium carbonate, which seals small cracks and reduces overall seepage through the RCC. However, two negative side effects result from this chemical process.
Relief wells drilled through the RCC can become restricted by the carbonate build-up — Willow Creek has had up to two inches of build-up on the galley walls, for example, and Quail Creek has already had its relief drains reamed out to remove the build-up. The other potential negative is the high pH, up to 11, of the seepage water. If seepage flows are large compared to other releases the receiving stream may suffer environmental damage. Grindstone Canyon experienced this problem and designed a system to pump the seepage water back into the reservoir.
Siltation within the reservoir can reduce seepage rates but this generally occurs after the project is well into its design life. Very few of the RCC projects in the USA have been in service long enough to experience a significant reduction in seepage rates due to siltation.
Performance monitoring has shown that the season can have a substantial bearing on seepage rates. Dams located in regions subjected to large seasonal temperature variations that have uncontrolled cracks will show more seepage through those cracks during the colder months: the dam contracts, opening up the cracks. When the warmer months approach the cracks will start to close.
Most cracking in RCC dams can be attributed to abrupt changes in the dam’s foundation profile and/or material properties, or to thermally induced stresses. These cracks are generally vertical and transverse to the dam’s axis, and pose no problems to the structural integrity of the dam. However, cracks can be a source of concentrated leakage; they can lose revenue for water supply reservoirs and create the appearance of a safety problem.
Designers of many earlier projects elected not to create joints in the dams or had wide spacing between joints with the understanding that any large uncontrolled cracks would be repaired. This earlier philosophy has changed: in most designs uncontrolled cracking is prevented.
Considerable insight has been gained into why cracking occurs and the analytical tools to evaluate cracking potential have improved to a point that crack spacing and widths can be fairly accurately estimated. The most common cause of cracking in RCC dams is thermal stress, which develops as a result of the differential in temperature — within the RCC mass and the exposed face, and between the mass and the ambient air temperature. If the dam is restrained from moving, to accommodate the strains produced by the stresses, cracks will develop if the thermal stresses exceed the tensile strength of the RCC.
Most all new gravity RCC dams higher than 50ft that are expected to impound a normal pool will have a thermal analysis performed as part of the design so that uncontrolled cracking can be minimised. There are generally two levels of analysis available to designers, with the selection based on the size and complexity of the dam. The level one thermal study is a mass gradient analysis which generally relies on material properties from published data and not on specific laboratory tests. Typical conclusions are: •Maximum RCC placement temperatures for different seasons.
•Crack spacing with total crack width potential.
•Sensitivity of the analysis to changes in input parameters.
More complex analysis creates a finite element model where site-specific material properties are used, detailed weather data is incorporated, and the exact geometry of the dam is input. Designers using finite element analysis, with accurate input data, have been able to reduce the occurrence of uncontrolled cracking.With a trained engineer, these models can and have been transported to the field and updated daily as the RCC placement proceeds.
Each day’s RCC and ambient temperatures are added to the model and if situations arise where placement temperatures or the ambient temperatures change significantly from the original model, modifications to the number of joints or their spacing can be promptly incorporated.
Crack control of cast-in-place concrete facing elements are now generally handled by constructing waterstop contraction joints around every 20ft. Some of the earlier dams used waterstops in the facing only at locations that coincided with joints through the dam and used bevelled notches as crack inducers for aesthetics. If a very workable RCC, and bedding mix, were not used the cracks in the facing allowed water to enter into the RCC between the lifts.
To reduce the cracking potential of RCC dams designers have developed controls on the placement temperature of the RCC mix, limitations on the seasonal placement window, and have investigated the properties of the individual mix materials. The target is to keep the peak internal temperature of the RCC mass as close as possible to the air temperature — the maximum temperature differential should not generally exceed 25-30degF.
Most methods of controlling the internal temperature of the RCC, as the cement hydrates, are confined to cooling the RCC material before placement and selecting materials that have a low heat of hydration. Cooling has been provided by evaporate cooling of the aggregate stockpiles, using chilled water of ice flakes for the mixing water, shading the aggregate stockpiles and cement silos and injecting liquid nitrogen into the mix.
Theoretically, liquid nitrogen and shaved ice provide the greatest cooling benefit, but there can be problems: the liquid nitrogen must have ample retention time in the mix, and the ice must be metered into the mixing chamber and melted before compaction.
A low-heat type II Portland cement, with a maximum hydration of 70Cal/g is the usual requirement: in some locations this requirement cannot be met or there will be a premium added to the cost of the cement. Designers must investigate whether the available cement has properties that will satisfy their design requirements.