The Ertan hydroelectric project is on the Yalong river in China’s Sichuan province. With an installed capacity of 3300MW, it is the largest hydroelectric plant in Asia to be completed this century. The main project components are a concrete double-curvature arch dam 240m high and 775m long and a huge underground powerhouse complex comprising a powerhouse cavern (281 x 26-31 x 66m) with six 550MW units, a transformer cavern (215 x 19 x 25m) and a surge chamber (201 x 19 x 69m).

The project also has: two diversion tunnels, presently the largest in the world (1000 x 20 x 23m); two spillway tunnels (each 850 x 13 x 13.5m); and a log passing facility with an annual mechanical log-passing capacity of 1.1 million m3 consisting of an 1800m long tunnel (17.5 x 7.6m) and a 500m long bridge. The contract for all underground works and related structures was awarded to a German-Chinese joint venture under the leadership of Philipp Holzmann, Germany. The other partners are Hochtief, Germany and Changjiang Gezhouba Engineering Bureau, China.

The powerhouse complex is 200-400m beneath the left bank slope of the Yalong river in a Permian basalt massif, which was intruded by granodiorite, syenite and gabbro during the Himalayan orogeny. The intrusive bodies are oriented NNE- SSW, as are the major tectonic fault zones, of which some still show seismic activity. During the pre-tender phase site inves-tigations were carried out. They were:

•371 bore holes totalling 30.4km in length.

•45 exploratory adits with a total length of 9km.

•Numerous in-situ and laboratory rock mechanics tests.

The cavern complex is mainly located in granodiorite and gabbro classified as good to fair rock, according to the NGI rock mass classification system. In-situ stress measurements revealed a major principal stress of 20-40MPa dipping at less than 30° towards the SSW. This is roughly three times the value of the effective minor principal stress or the vertical strain resulting from the actual rock mass cover.

The designed rock support for all tunnels, caverns and shafts consisted of untensioned, fully grouted, 4-8m long rockbolts without endplates, to be installed as initial support. After the completion of all rock bolting work, layers of plain or wire-mesh reinforced shotcrete were to be applied, according to local requirements. In a third step, rows of prestressed anchors 20-25m long, with a working load of 1750kN, were planned in the side-walls of the powerhouse cavern and the surge chamber.

Excavation experience

During the excavation of access and pilot tunnels into the powerhouse cavern complex several violent rockbursts took place in areas of massive granodiorite and gabbro, causing fatalities and damaging equipment. The time and place of the events were unpredictable and they caused damage despite the supports put in place in the design.

Austria’s Geoconsult was invited to simulate and analyse the rock mass and stress conditions and to give advice on the most suitable excavation and support measures. An elastic 3D FE analysis (see top diagram opposite) shows two major stress concentration zones in each of the three caverns. One zone is in the upper western side-wall and lower western crown area, and the other is in the lower eastern corner of side-wall and invert. The latter zone moves downwards with the excavation level as the bench excavation proceeds. This means that the entire eastern side-walls of all three caverns could be affected by stress related rock failures. Maximum compressive stresses on the order of 50-65 MPa could be identified in both stress concentration zones. After these high-stress zones were identified the locations of rockburst events were more predictable and a support concept could be worked out to reinforce them. However, the main problem continued to be the timely installation of effective support to prevent serious accidents. The time between blasting and rock burst occurrences was in general too short to install the complete design support as protection against such events.

Most accidents occurred during rockbolt installation, when personnel and equipment were completely exposed to the violent expulsion of rock fragments. Even the installation of shorter rockbolts in a denser pattern did not improve the working conditions sufficiently.

A significant improvement in working safety was the introduction of steel-fibre shotcrete as a new support element. It possessed sufficient strength and ductility to withstand the high energies released during rockburst activity and was also acceptable as a final support element. However, to be effective as a protective measure, steel-fibre shotcrete had to be applied prior to rock bolting and rock bolts had to be installed with endplates.

Instabilities and overbreak

During bench excavation, 2m wide rock seams were left in place adjacent to the side-walls for profiling by smooth blasting. However, at the eastern side-walls, spalling and buckling of rock slabs occurred prior to profiling, and stress related fissures developed with dip angles of 50° towards the opening. These fissures reached into the side-walls and caused considerable overbreak during profiling. These stress-related phenomena occurred mainly at the bench toe of the eastern side-walls, as the model study predicted. Consequently, due to the local stress conditions, the same blasting and excavation method would often produce a neat profile with all the visible half barrels of the blast holes at the western side-walls and an irregular profile with much overbreak at the eastern side-walls.

During the excavation of the crane rail berms in the powerhouse cavern, the combination of stress-induced fissures and existing joint sets created large overbreak volumes, which could not be prevented even by very careful phased blasting and protective bolting.

The edges of the berms were in general unstable and broke off during blasting, or were so loose that they had to be removed by scaling. The missing berm portions had to be re-established with concrete.

Rockburst damage

During benching, a series of rockbursts damaged the final lining of wire-mesh reinforced shotcrete in the western springlines of all three caverns. Steel-fibre shotcrete was applied and additional rockbolts with large endplates were installed in the damaged areas as remedial measures above the ongoing excavation.

When the power conduits and bus tunnels were excavated, the frequency and intensity of rockburst events increased and more areas were damaged. Bus tunnels 1 and 2, located in massive granodiorite, were frequently affected by severe rockbursts, causing several cave-ins. After each event the support was reinforced by additional rockbolts and steel-fibre shotcrete layers. In spite of these measures, new cracks developed inside the bus tunnels and in the western side-wall of the transformer cavern.

The culmination was a severe rockburst with its focus in the powerhouse bench invert. The event caused cave-ins at the western springline of the transformer cavern, from where patches of shotcrete and rock fragments were expelled up to the eastern side-wall of the cavern. Bus tunnel 2 was so badly damaged that it was not accessible for weeks. In the powerhouse cavern, new cracks extended over 40m in the western springline, where additional steel-fibre shotcrete and additional short rockbolts with large endplates had been installed as remedial works.

The result of these events was progressive deterioration of the installed support system, in areas where excavation and support installation had been completed several months or even years before.

The modifications in the support system, which had dampened the severity of rockbursts for a long time, could no longer provide safe working conditions.

Following these events, the joint venture requested elaboration of a more complex stress model to simulate the effects of secondary structures on the stress field. It revealed additional stress concentration zones near the intersections of structures, along the bus tunnels and draft tubes, in the end walls of the caverns, in the rock pillar between surge chambers 1 and 2 and in the turbine pits of the powerhouse cavern. In the latter areas stresses up to 120 MPa were predicted.

As a consequence of the new findings end-anchored cone bolts and partially grouted rockbolts with large triangular endplates were introduced as additional support elements in the newly defined critical zones. These types of bolts can accommodate large deformations.

Major events

The most violent rockburst happened in September 1995 — fortunately during a break, when all the personnel had left the site. It caused severe damage in the western side-wall of the transformer cavern, the eastern side-wall of the powerhouse, in bus tunnels 1, 2 and 3 and in draft tubes 1 and 2.

An analysis of deformation measurements and of visible damage suggests that the rockburst was caused by a complex block movement towards the powerhouse cavern along intersecting shear zones, which were reactivated by the process of stress redistribution.

In April 1996 another major rockburst event happened in the powerhouse invert between turbine pits 1 and 2. The rockburst activity lasted for three hours and caused uplift and fragmentation of massive rock layers to a depth of 6m below the invert and the lower eastern side-wall of the powerhouse cavern.

The event triggered spalling in the western springline and the failure of eight prestressed anchors in the eastern powerhouse side wall between bus tunnels 2 and 3. Several concrete columns, which supported the suspended ceiling, showed vertical cracks at the contact with the side-wall.

Longitudinal hairline fissures in the suspended ceiling and deformed GEWI bars indicated that large side-wall displacements had caused bending and lifting of the suspended ceiling.

Deformation measurements showed considerable displacement in the lower eastern side wall of the powerhouse cavern, where block movement had taken place during the previous rockburst.

Extensometer measurements showed very high side-wall deformations in all caverns. At the completion of all excavation works, maximum wall convergence was 121mm in the powerhouse cavern, 184.6mm in the transformer cavern and 190.6mm in the surge chamber. Time-deformation graphs showed the deformation increased step-wise, related to the different excavation levels and the major rockbursts.

These deformations were much higher than expected and completely unacceptable according to the criteria of Fujita et al (Proc 1st Int Symp Rockstore 77) and Kaiser (Int Symp Large Rock Caverns, Helsinki, 1986), who report 50mm and 25mm, respectively, as admissible wall deformations.

Other authors consider cavern deformations exceeding 0.1% of the span as not recommendable in view of admissible field strains. Maximum deformations were 0.4% of the largest span in the powerhouse cavern and 0.96% in the surge chamber.

Remedial work

Despite all these support modifications and reinforcements, there was progressive cracking of the shotcrete lining in all stress concentration zones after the reinforcement of the lining. Several rockbursts showed that even a steel-fibre shotcrete lining 15-20 cm thick and a closely spaced pattern of rockbolts provided no guarantee of structural stability and working safety.

To mitigate any further interruptions to the works due to unsafe working conditions, all the damaged areas were covered with galvanised chainlink mesh fastened to existing and additional partly grouted rockbolts. This works was very difficult, because it had to be performed up to 50 m above the ground. In the surge chamber temporary rail beams had to be built onto protruding anchor blocks in the upper side-walls and the work was carried out from a working platform on rails. Since the use of chainlink mesh was very effective in preventing rock fall incidents, it was generally adopted as remedial measure in most of the areas with stress related damages.

The mesh was permanently secured by a systematic pattern of partly grouted single strand cable anchors, tensioned to 150kN. In the crown of the powerhouse a continuous steel cable was fastened to the anchor plates in a zigzag pattern below the chainlink mesh to protect the suspended ceiling against damage from potential cave-ins.

Additional multiple tendon prestressed anchors had to be installed in the highly over-stressed pillars between the caverns and in the invert of the powerhouse cavern. These anchors were not tensioned to their full working load in order to avoid failure due to further rock deformation. Radial beams of reinforced shotcrete and single strand cable bolts were installed as reinforcement of the damaged draft tubes 1 and 2.

It is obvious from the tangential stresses of 50-120 MPa that no support system can provide sufficient confining stress to prevent stress related rock failures. During the entire construction time at Ertan the joint venture had to deal with new problematic situations. The experience showed, however, that long prestressed anchors in a dense pattern were effective in limiting the damages of rock bursts. Their function might have been limited during the initial excavation phases, when deformations and stress related rock failures were rather shallow. But as deeper regions were affected by stress redistribution and larger blocks were mobilised by severe rock burst events, prestressed anchors were an important retaining element. The cavern support was improved by changing the sequence of installation and introducing ductile and yielding support elements, such as steel-fibre shotcrete, end-anchored bolts and single-strand cable anchors. Using galvanised chainlink mesh helped improve working safety.

Flexibility and improvisation was required to implement remedial and support works, under difficult circum-stances, without exceeding the various milestones for completion of the works.