Dam Safety: Investigating Failures of Post-Tensioned Anchors

This article has been evaluated and edited in accordance with reviews conducted by two or more professionals who have relevant expertise. These peer reviewers judge manuscripts for technical accuracy, usefulness, and overall importance within the hydroelectric industry.

By Malte O. Cederstrom, Vattenfall

Vattenfall AB, Sweden's largest power producer and Europe's fifth largest, operates 53 large hydropower plants in Sweden. The 33 terawatt-hours of hydroelectricity generated by these plants, along with Vattenfall´s nuclear power, account for about half of the country's power production. Vattenfall places great importance on its dam safety activities, which include research, emergency preparedness, surveillance, safety evaluations, and system-wide programs of safety upgrades.

Several of Vattenfall's concrete dams are stabilized with post-tensioned anchors. Surveillance of these dams includes a regular check on the functionality of the anchors. In 2002, a load test of anchors at the 120-MW Alvkarleby hydro project on the Dalalven River, 170 kilometers north of Stockholm, revealed that seven of the 78 anchors in the intake canal wall had ruptured. Diagnosing the cause of the failures proved to be a lengthy process involving the design engineer, the manufacturer, the installer, and laboratory specialists performing chemical and metallurgical analysis. As a result of the investigation, Vattenfall installed a different type of anchor throughout the facility and no longer relies on the remaining anchors.

Installing the anchors at Alvkarleby

The Alvkarleby plant was constructed in the beginning of the 20th century and began producing power in 1915. The original plant had five generating units and a total capacity of 70 MW. The development included a 200-meter-long intake canal partially blasted in rock and partially lined with concrete walls. The concrete walls were cast in 10- to-15-meter-wide monoliths ranging in height from 3 to 16 meters. Drainage pipes were installed in the walls, and the expansion joints were sealed to prevent leakage.

In the late 1980s, Vattenfall added a new 50-MW unit and refurbished the old units. To accommodate the increased turbine flow, the intake walls were raised and reinforced. Design loads considered in raising the walls included water pressure, uplift, ice, and load rejection. The left wall was raised about 0.6 meter, and a 0.3-meter thickness of concrete was cast on the inside of the wall, with reinforcement bars connecting the new concrete, old concrete, and rock.

To further stabilize the wall, Vattenfall contracted with a construction company to install 78 36-millimeter post-tensioned anchors in holes drilled through the old concrete and 6 to 8 meters of the underlying rock (see Figure 1). The holes were tested for watertightness before installation of the anchors and, if leaks were detected, were pressure-filled with grout and redrilled. After refilling the holes with grout, the contractor fixed the anchors in the holes, grouted the 5-meter anchoring zone, and allowed the grout to cure. The upper anchor plate was then grouted into place and allowed to cure, and the anchor was tensioned to a force of 720 kiloNewtons. The tension force applied to the anchors was 66 percent of the nominal yield strength and 58 percent of the nominal ultimate strength for the anchor material. During installation, the anchors' elongation was also measured at various loads and compared with expected values. Before the installation, the anchors were wrapped in a corrosion-protecting band that would allow some movement inside the concrete. This would enable Vattenfall's dam safety engineers to measure the actual force in each anchor in the future. Finally, the hole was grouted and the top of the anchor was treated for corrosion protection and capped.

Discovery and initial investigations of failures

After installation, Vattenfall established a program to sample the anchors' performance at five-year intervals. Tests consisted of measuring the force needed to just release the nut. In 2002, the test revealed seven broken anchors. At the same time, the engineers realized that the anchors had been installed with plastic caps, which needed to be replaced with steel. They also noted that the top corrosion protection seemed to be of poor quality. Over the following two years, three more anchors ruptured; two during a test. All of the fractures developed less than 2 meters from the anchor's top, and most less than 1 meter from it.

Click to Enlarge
The Alvkarleby hydroelectric plant has been in service since 1915. The intake canal wall in which the anchor failures occurred is behind the bridge.

After the initial discovery, Vattenfall contacted the design consultant, contractor who delivered and installed the anchors, and supplier. The installation contractor sent two of the failed anchors to an independent metallurgical laboratory, where they were examined microscopically and subjected to strength tests. The fracture surfaces were corroded, making visual assessment difficult. The laboratory reported that the fracture surface appeared both ductile and trans-crystalline and that the fracture appeared to have been instantaneous. On one surface, a crescent-shaped initiation point was evident. The yield point and rupture strength in the material were 3 to 5 percent lower than the European standard values of 1,080 megapascals and 1,230 megapascals, respectively. The laboratory report indicated that the probable cause of failure was an overload, such as a sudden high water level.

The design consultant, who was not aware of the laboratory's findings, also suggested a sudden water level increase as a possible cause, along with ice loading, bending of the anchor at the plate, and material weakness in the anchor bar. Finally, the anchor supplier mentioned that welding often leads to ruptures in this type of high-strength material.

Pursuing a better explanation

After reviewing the initial reports, Vattenfall's dam safety engineers decided a more detailed investigation was needed in order to prevent similar incidents in the future. The team of four engineers assigned to the investigation brought expertise in metallurgy, concrete, design, and dam safety. In conducting the investigation, they used Vattenfall's metallurgical and concrete laboratories in Alvkarleby, as well as other laboratories. The investigations included chemical and metallographic analyses, magnetic particle testing, penetrant testing, ultrasonic inspection, impact testing, and grease analysis.

Chemical and metallographic analyses
Ruptured anchors were subjected to a chemical analysis using optical emission spectroscopy. The analysis indicated that the constituent elements in the steel fell within published industry standards, with the exception of a slightly lower-than-standard manganese content. The microstructure of one of the anchors also was studied under an optical microscope and was consistent with the chemical analysis.

Vattenfall's laboratory staff also performed nondestructive testing on the ruptured anchors, including ultrasonic testing of the end of the bars and dye penetration tests and magnetic particle inspection of the surface of the ruptured anchors. None of these tests revealed cracks in the anchors. At the same time, laboratory staff performed a separate test of the ultrasonic method on an intentionally cracked bar. The test showed that the method did not have the sensitivity to detect flaws of the size believed to have initiated the ruptures.

To evaluate the brittleness of the steel and the transition temperature, the laboratory also performed a notched-bar impact test on samples from one of the anchors at temperatures from -10 degrees Celsius (C) to 25 degrees C. The resulting test values were 2 to 3 Joules, much lower than the expected value of at least 30 Joules for common construction steel alloys. The low test values indicate a high degree of brittleness.

Grease analysis
During installation, the anchors were covered by a special anti-corrosive grease and wrapping, in addition to the corrosion protection provided by the grout. Oxidation of the grease would increase the amount of free acid present. Grease samples from three failed anchors were analyzed and found to be more acidic than a sample of unused grease provided by the supplier. However, the test values were not consistent enough or high enough to support the theory of embrittlement due to deterioration of the grease.

Load tests
To verify the strength tests made on small specimens in the initial phase of the investigation, a laboratory was engaged to perform uniaxial load tests on two of the ruptured anchors. Rupture occurred at 1,159 and 1,136 kiloNewtons, or 7 and 10 percent lower than the European standard. The characteristics of the fracture surfaces produced during the tests were similar to those observed on the in situ fractures. On at least one of the samples, a crescent-shaped initiation zone, similar to the in situ fractures, was visible.

Investigating a fracture surface in detail

The anchor having the fracture surface least affected by corrosion was chosen for detailed microscopic investigation. A stereomicroscope photograph clearly showed the crescent-shaped. 3.2-millimeter by 0.8-millimeter initiation zone, as well as rust stains. The initiation zone showed signs of oxidation and was sharply delineated from the remainder of the fracture surface. The sharp difference between the initiation zone and the remaining area allowed the investigators to rule out fatigue, which would have resulted in crack propagation lines. Failure due to overloading would have been marked by deformation, which was not detectable in the sample investigated. Finally, the absence of visible grain boundaries in the area of the fracture showed that coating by brittle phases in the grain boundaries, such as cementite or sulphide of iron, was not the cause.

A more detailed microscopic view showed significant details of the boundary between the initiation zone and the remainder of the fracture surface. A structure just inside the initial fracture zone contained needle-shaped separated crystals, such as would be formed by precipitation from a liquid or gas phase.

These crystals could not have been formed during the steel manufacture process, because the subsequent hot-rolling of the bar would destroy them. The investigators concluded that the crystals were the result of water entering the crack, and therefore that the crack formed after the anchor came into use.

Reviewing possible sources of failure

The various professionals queried about the failures raised several possible explanations, including ice loading, transient water pressures due to a plant trip, bending, and damage during welding. Vattenfall's investigative team considered each of these but did not find a persuasive case for any of them, based on the laboratory results and other project data. Overloads, due to either ice pressure or plant tripping, should have left evidence in the form of deformation in the fracture surfaces. Vattenfall also had performed calculations for ice loading and physical model and full-scale tests for loading during a plant trip and found that the resulting loads did not exceed the actual force of the tensioned anchors. Regarding ice loads, at least some of the ruptures occurred during ice-free periods. Ice formation is normally prevented in the intake canal, although no records existed to confirm this for the period of outage for refurbishment.

Click to Enlarge

Anchor bars also may be subject to bending, caused for example by an anchor plate that is installed at an incorrect angle. However, the spherical surface of the nuts used should have prevented bending by this mechanism. In addition, the ruptures were too far from the nuts to support this explanation.

Welding damage also was considered an unlikely explanation, since there was no record of welding at the time of installation. The protective wrapping should also have protected the anchors against welding sparks, and the fractured surfaces showed no sign of welding damage.

It was apparent that the ruptures had occurred well after the installation and tensioning of the anchors. Generally, explanations for this type of failure could be either fatigue or environmentally induced cracking of the material under stress. The investigative team rejected the fatigue hypothesis because the loads acting on the anchors did not vary significantly over time and because of the absence of stop lines on the crack surface.

Pinpointing the cause

As a result of the investigations, particularly the investigation of the fracture surface under a scanning electron microscope, the team concluded that the most likely cause of the failure was environmentally assisted cracking. Environmentally assisted cracking could be brought about by either hydrogen-induced brittleness or stress corrosion. The hydrogen theory was based on the fact that, over time, additives to the concrete grout develop free hydrogen, which has a high rate of diffusion into concrete and steel. Increased hydrogen content in a high-carbon steel could cause the steel to rupture. The time to rupture would depend on the kind of steel, hydrogen content, and stresses. However, the grease analysis had not shown significantly elevated levels of free hydrogen. Therefore, the investigators believed that the anchors had not undergone prolonged exposure to free hydrogen. Furthermore, hydrogen embrittlement would not have been limited to the top 2 meters of the anchors, where all of the ruptures occurred.

The explanation for the ruptures finally adopted by the investigative team was stress-induced corrosion, in combination with the high degree of brittleness of the steel. The visible oxidation in the initiation area pointed to corrosion, and the brittleness and relatively low steel strength explained the material's low resistance to stress corrosion. The anchors initially had been installed with a plastic cap to keep grease in and water out. However, the caps proved to provide inadequate corrosion protection. Inadequate protection at the top of the anchors might explain the location of the ruptures in the top 2 meters of each anchor.

Lessons learned from the investigations

As a result of the anchor failures and subsequent investigations, Vattenfall installed steel cables to replace all the anchors at Alvkarleby. The remaining anchors were left in place but were assumed ineffective in stability calculations. Vattenfall now checks the few anchors of the same type installed in its other dams by pulling with a jack until the nut is just released.

The investigations also brought new insights about methods of testing anchors in dams. When the ruptures were discovered, Vattenfall was testing anchors by measuring the actual load in a sample of anchors with a hydraulic jack. A solution used at some other plants was to install additional test anchors in the dam, which were used for testing but not considered in stability evaluations or design. Vattenfall also experimented with ultrasonic testing but found in the Alvkarleby tests that the method could not detect some small but potentially serious fractures. Vattenfall has joined a research effort with CEATI Dam Safety Interest Group to find new and better methods for anchor testing.

Correcting the situation

At the conclusion of the investigations, Vattenfall's dam safety engineers recognized that the ruptured anchors should not be replaced in kind and the remaining anchors should not be relied upon for stability. Three main alternatives were considered for meeting safety criteria: installing new anchors and strengthening the existing concrete; constructing supporting walls; and complete replacement of the dam. Although there was a strong interest within the company in alternatives that did not involve anchors, the first option was clearly the most economical. The supporting walls were aesthetically undesirable, particularly because the Alvkarleby plant is a tourist attraction and is known as a facility with few negative environmental effects. There was also little space for such supports. The cost of a complete dam replacement was much higher than that of the other alternatives, especially considering the cost of lost generation.

Concrete test results from the investigation showed that the existing concrete in the dam was in poor to very poor condition. The situation provided an opportunity to apply the results of one of Vattenfall's recent research efforts. This research project had focused on using concrete grout injections to increase the life of aging structures. The grout composition and the specific installation techniques used at Alvkarleby were based on lessons learned through the research. Grout was injected into holes drilled at 1-meter intervals along the dam. After completing the grouting, the contractor drilled holes for the new anchors and the drill cores were saved for analysis. The analysis of the cores showed that the grout injection had substantially improved the condition of the concrete.

The new anchors comprised 40 steel tendons, each consisting of 12 12-millimeter-diameter wires. The wires were split apart in the lowest 5 meters to improve the anchoring. Above the anchor zone, the wires were coated in protective grease and placed inside plastic pipes that would allow the wire to move freely when tensioned. The anchor holes were filled with grout and the tendons lowered into the grout. After the grout had cured, the anchors were tensioned to 1,500 kiloNewtons.

The ruptured anchors were replaced in 2004, and the remaining new cables were installed in 2005. Vattenfall plans to select a sample group of anchors for testing at five-year intervals. The expected life of the new anchors is about 50 years, and there is ample space to install new ones when needed.


Cederstrom, Malte, Per-Erik Thorsall, Bengt Hildenwall, and Stig-Bjorn Westberg, "Incident with Loss of Seven Post-Tensioned 72 Ton Anchors in a Dam," Dam Safety 2005 Proceedings, Association of State Dam Safety Officials, Lexington, Ky., 2005.

Malte Cederstrom is a senior civil engineer in Vattenfall's hydropower division. He was responsible for directing the investigations of the failed anchors and developing recommendations for their replacement. 

More Hydro Review Worldwide Issue Articles

To access this Article, go to: