Alkali aggregate reaction (AAR) in concrete proceeds over decades and occurs in all types of concrete structures, but its results – slow but unstoppable growth – are especially important in dams because of their longevity, the sheer bulk of concrete that can be affected, and the disastrous effects of failure. What is more, the requirement to maintain mechanical equipment buried within the concrete to precise specifications presents a deeper challenge than observing the dam mass alone, and it is a challenge where remedial work is expensive.

The cost of dealing with AAR was clearly laid out at the four-yearly International Conference on Alkali Aggregate Reaction in Concrete, which was held in June 2000 in Quebec, Canada. For example, Marie Gaudreault of the St Lawrence Seaway Management Corporation discussed regular work carried out at four locks on the Seaway.

At these locks, due to AAR identified in the 1970s, stop log guides and ice gates were squeezed; pipes and shafts, and gates and valves, had become misaligned; there was cracking within the cable galleries and the coping concrete had deteriorated. The affected mechanical structures – particularly the gates and valves – have to be regularly repositioned and realigned, along with all their fixings and parts of the concrete surround. The total cost per lock is well over US$650,000 and the interventions must be repeated every 10 to 20 years. Modifying the concrete recesses for mechanical components that close across the lock costs from US$787,000 to US$1M per lock, depending on the number of recesses. A grouting programme is necessary to stop water infiltration reaching mechanical and electrical components, even though it is well known that the grouting can only be a temporary measure, as the cracks will continue to open. The current grouting programme will cost at least US$164,000 annually over the next six years.

An example of remedial work on a dam was discussed in detail by Dan Curtis of Acres International. He examined the case of the Mactaquac dam on Canada’s St John river. Owned by New Brunswick Power, the dam and its six generating units went into operation between 1964 and 1968. The rockfill dam is 518m long and 46m high. The dam showed evidence of AAR by the mid 1970s, and a longitudinal vertical construction joint was seen to be expanding. By the early 1980s, leakage was established through expanding construction joints in the spillway, intake and diversion sluiceways.

In response, slots were cut in the intake with a 13mm diamond wire saw in 1988, 1989 and 1992. It is expected that more slot-cutting will be required in future.

In the power house, concrete expansion and substructure movements were inducing extra load on the draft tube piers and superstructure frames; the generating units were suffering ovalling of the discharge ring and stay ring distortion; and unit alignment was under pressure. In 1995, after analysis using a finite element program called Grow3D, seven transverse slots were cut in the power house between the units. As anticipated, the slots were followed by a rebound of 7mm and were then closed at a rate of 2mm per year. The slots were due to be recut in 1999 and 2000.

Cuts were also made in the penstock encasement just upstream of the power house to relieve distortion of the turbine stay-rings, following a numerical analysis using Grow3D. From 1996, the penstocks were cut at a rate of one per year.

Similarly, the Paulo Afonso IV dam, on Brazil’s Sao Fransisco river, was built in the late 1970s and its six power houses were all in operation by May 1983. Evidence of AAR was identified from 1986, when cracking was observed in concrete structures near the generators. Some mechanical problems were also observed, including:

  • Inclination of the turbine generator axle.

  • Tilting of the turbine top.

  • Variations in the tolerance space between the paddles.

  • Deformation of the turbine case.

In 1995, the owner of Paulo Afonso, Companhia Hidro Eletrica do Sao Fransisco, began to study potential remedial measures for the expansion effects. Following the work carried out at Mactaquac, cutting expansion joints is high on the list of work. In one scenario, a cut would be made to form transverse expansion joints between the generating units and a longitudinal joint along the interface between the concrete housing the generating units and downstream of the cavern. Alternatively, a longitudinal joint downstream, between the generating units and the rock mass, is also under investigation. But the company is aware that cutting expansion joints is not without its own problems: relieving stresses in this way intensifies concrete expansion in the immediate region.

Slot cutting has been hotly debated, and its demerits were discussed in a paper by V Gocevski of hydro-quebec and S Pietruszczak of McMasters University. These authors argue that few numerical models have been available and there has been little analysis of the effect of slot-cutting, so that in the past it has often been carried out on little more than the intuition of the designer.

Pattern of natural cracking

The authors say that the magnitude of the distortions formed are small in respect of such massive gravity structures, and that the pattern of natural cracking that forms is the least forceful way to accommodate the movements caused by the expansion. Citing the results of pre and post slot-cutting at Beauharnois and Mactaquac, they note that the stresses caused by the growth were in the region of 3.5MPa and slot cutting reduced this by 0.9-1.3MPa – both far below the recognised compressive strength of concrete, which is some 30MPa.

Transverse slot cutting between generators is intended to reduce stresses uniformly and address the problem of maintaining clearances in the turbines. The authors argue that since the concrete masses above the spiral cases are not symmetrical, such cutting cannot have a uniform effect, and say that this is borne out by post-cutting monitoring at Beauharnois. Gocevski and Pietruszczak argue that minor interventions should be sufficient to deal with continuing AAR expansion. It should be noted that the effects of AAR vary depending on the dam structure. Dan Curtis also reported on several arch dams including Alto Ceira and Santa Luzia in Portugal, Cahora Bassa in Mozambique, Kouga in South Africa, and Gene Wash and Copper basin in the US, and he presented a general analysis of the stresses placed on arch dams by AAR. He used Grow3D to examine stresses within arch dams, the results of which showed that expansion due to AAR tended to shift the dam upstream, increasing compression on the upstream face (which acted to reduce tension near the base of the dams on the upstream face that could exist with AAR loading) and reducing compression on the downstream face. On most of the downstream face, large tension forces did not develop as they were countered by the increased rate of concrete growth. The area of greatest tension was the upper part of the downstream face, and as the central portion of the dam rose it could cause tension stress near the abutments.

Reports from Brazil’s Peti dam, completed in 1946, bore out these conclusions. Cracking was first observed near the left abutment in the 1960s. The area has been patched but cracking has continued to develop since then on a continuous basis and has weakened the area around the abutment. The operator is monitoring the effects on a three-year basis.

Source of the problem

If remediation work on dams suffering from concrete growth is one part of a strategy to deal with the problem of AAR, equally important are measures to ensure that the problem does not arise in new structures.

Dealing with AAR at source is made more difficult by the fact that the chemistry involved in the reaction is not well understood. In simple terms it is an attack by sodium or potassium hydroxide solution, arising from the cement, on silica in the material. An alkali silica gel is produced which absorbs water, and as water is absorbed the gel expands, causing the concrete growth.

What is certain is that the susceptibility of concrete to AAR depends very much on the type of aggregates used – granites and gneisses, for example, are susceptible. The UK Concrete Society’s guidance notes state that ‘damage caused by AAR can only occur if all three of the following factors are present in the concrete: sufficiently high alkalinity; a critical amount of reactive silica in the concrete; and sufficient moisture’. But the minimum conditions for damage may tell us little about the process by which the gel grows. The society says that there is ‘no easily identifiable relationship between the abundance of gel deposits and the magnitude of any resultant expansion or cracking’ and notes that in the UK it is common for damage to have occurred when the reaction appears, in thin sections, to have generated relatively small amounts of gel.

Guidelines in the UK list the types of material available likely to provide the aggregates for new concrete construction and their susceptibility to AAR. But in the UK, as Rebecca Hooper from the Building Research Establishment explains, engineers have a certain amount of choice: the UK is a small country with good transport links, and a wide variety of aggregates are available. The problem in dam construction is often that many thousands of tonnes of aggregate are required, and they are often built in mountainous or other inaccessible areas where it is simply not possible to bring in aggregates from elsewhere. In that situation, the local aggregates are the only practicable option. Add to this the second condition for AAR development – the constant presence of water – and it can easily be seen that dam construction is particularly at risk of AAR development.

Preparing to deal with the problem begins with a full understanding of how susceptible local aggregates are to developing AAR. Individual countries have used tests on concretes for years but in May this year internationally accepted standard tests were introduced, developed by a technical committee of the International Union of Testing and Research Laboratories for Materials and Structures (Reunion Internationale des Laboratoires d’Essais et de Recherches sur les Materiaux et les Constructions, known as RILEM).

Dr Ian Sims of the UK’s STATS, a specialist engineering, materials and environmental consultancy, is secretary to the technical committee (designated TC106). ‘Reliable tests can manage the problem of AAR,’ says Sims. ‘With only a few exceptions you can take remedial measures if you know you have a reactive aggregate.’

With more than 50 members, the technical committee represents a truly international attempt to develop consistent tests that can be used worldwide: important not only to allow effects to be compared internationally, but also to allow companies to transfer their experience when they work in different countries.

‘Every country had its own tests,’ says Sims, ‘and this was a chance to find universal tests that could be applied worldwide and to any materials.’

In developing standards, ‘We found a few tests were local and could not be used more widely,’ Sims continues. ‘Also, with some tests, it is difficult to relate the test to what really happens.’

It was important to use tests that produced results quickly. Tests often took a year or more and by then the project might be completed. ‘The problem was that too many tests were too long term,’ Sims adds. ‘What we want is for the test to relate to reality, and then to be accelerated.’

The new tests will allow a systematic assessment to be carried out not only locally, but in a form that can be translated worldwide. With a detailed knowledge of the reactivity of local aggregates, engineers can alter the concrete mix and introduce admixtures that will reduce alkalinity and slow AAR development.

Alleviate the burden

Several admixtures are in common use. For example, in the UK and in some other European countries, fly ash is a common constituent of concrete not only as a way of disposing of this waste material that is produced from burning coal, but also because its chemical composition acts against the development of AAR. In the UK fly ash qualified to a British Standard has been used in this way for many years, but it has not been so elsewhere.

That may change in the next couple of years, as a new European standard on concrete (EN206) is due to become law across the continent at the beginning of 2003. Before this the European standard on fly ash – EN450 – must be harmonised across Europe, and this may make the use of fly ash more common and help remove anomalies such as in France, where ash can be added to pre-mixed concrete at source but cannot be added if concrete is mixed on site.

Other concrete admixtures that may alleviate the burden of AAR include waste cement, silica fume and lithium in various proportions. In some cases, concrete using these admixtures is used in remedial work, for example at Australia’s Canning dam. Here some 4m of badly cracked concrete at the crest of the dam is due to be removed and replaced with reinforced concrete which will also act as a head beam, anchoring new post-tensioning which will be installed in the dam wall. Some 12 concrete mixtures were tested and the likely choice will contain 40% fly ash and 5% silica fume.

New tests available

Meanwhile, in Brazil, tests are continuing on using ‘rock flour’ as an addition to the concrete mix. Used at the Capanda dam in Angola, this powdered aggregate offers both a physical and chemical hindrance to AAR, and may also improve the workability of the cement. As well as decreasing alkali concentration, it partly fills the spaces between aggregate particles, reducing the sites available for the reaction to take place.

For existing structures, AAR is a continuing problem and requires careful monitoring and regular – and expensive – remediation work. But with new tests available to designers, and a variety of admixtures available to alleviate the high alkalinity of some aggregates, AAR is likely to be far less of a problem in new dams.