Tunnelling and Underground Space Technology 18 (2003) 223–232

Evaluation of loosened zones on excavation of a large underground rock

cavern and application of observational construction techniques
Toshio Maejimaa, Hiroshi Moriokaa,*, Takayuki Morib, Kenji Aokic
a
The Tokyo Electric Power Co. Inc, Tokyo, Japan
b
Kajima Corporation, Tokyo, Japan
c
Kyoto University, Kyoto, Japan

Abstract

To ensure the stability of caverns and optimize the support systems, observational design and construction procedures have
been used recently for excavating large underground caverns, such as those for pumped-storage power plants, which spontaneously
process data monitored during construction and real-time feedback of the results to the next construction steps. In observational
design and construction procedures, the cavern support systems should be optimized by monitoring the loosening behavior and
correctly predicting and analyzing the subsequent behavior of the cavern. This paper describes an evaluation of loosened zones
based on monitored stress of rock and an application of prediction analysis for observational construction techniques.
䊚 2003 Elsevier Science Ltd. All rights reserved.

Keywords: Underground cavern; Observational construction technique; Excavation; Support design; Monitoring technique; Numerical analysis

1. Introduction tional strain and rock displacement measurements to

directly evaluate the loosened zones. Prediction analyses
The Kan-nagawa power plant is a pumped storage
were performed by conducting precise preliminary geo-
power station of maximum output of 2820 MW and is
logical surveys and considering the results of inverse
being constructed by the Tokyo Electric Power Co. Inc.
analyses to improve the precision of prediction (Maeji-
150 km north-west of Tokyo at the boundary of the
ma et al., 2001a,b).
Gunma and Nagano Prefectures.
As shown in Fig. 1, the underground cavern of the
power station has dimensions of: length 216 m, width 2. Geology and rock properties
33 m, height 52 m, approximately 220 000 m3 in total
excavation volume, and maximum cross-sectional area The geology of the Kan-nagawa site was the Chichibu
of 1400 m2. south belt strata of Paleozoic–Mesozoic layer, which
The excavation was completed in October 2000. Four were formed by the re-deposit of marine sediment of
452 MW electric generators and two transformers will the Paleozoic–Mesozoic period during submarine land-
be installed in the cavern to generate an output of 1880 slides in the Jurassic–Cretaceous period.
MW. The strata consist of mudstone-based rock with irreg-
In observational design and construction procedures, ularly mixed olistoliths, such as sandstones, chert, basic
loosened zones around cavern during the excavation volcanic rocks, and limestones. As shown in Fig. 2, the
should be monitored and their behaviors should be strata near the power plant were classified into six by
correctly predicted to optimize the support systems. the matrix and the kinds and percentages of mixed
Rock strain measurements and acoustic emission gravel. The CG1 layer, which consists of conglomerate
(AE) measurements were conducted as well as conven- and coarse sandstone, was likely to be the prime bed.
The six layers were found to have different properties.
*Corresponding author. The Tokyo Electric Power Co. Inc, Con-
struction Dept., Ushisaiwai-cho 1-1-3, Chiyoda-ku, Tokyo, Japan 100- The rock property values and the initial rock stresses
0011. Tel.: q81-3-4216-1111; fax: q81-3-4216-4269. are shown in Table 1 and Fig. 3. Since steep discontin-
E-mail address: [email protected] (H. Morioka). uous surfaces were dominant in the M1, S1 and FS1

0886-7798/03/$ - see front matter 䊚 2003 Elsevier Science Ltd. All rights reserved.
PII: S 0 8 8 6 - 7 7 9 8 Ž 0 3 . 0 0 0 3 1 - 2
224 T. Maejima et al. / Tunnelling and Underground Space Technology 18 (2003) 223–232

Fig. 1. Dimensions of the underground power plant cavern.

layers, the elastic coefficients were anisotropic. Since that considered strain softening. A result of a prediction
the tailrace sidewall was to be built through the FS1 analysis is shown in Fig. 4.
layer, the behaviors of the FS1 layer had to be correctly The behaviors of the rocks at the site were character-
predicted and evaluated. A preliminary survey showed ized by:
local fluctuations in initial rock stress, showing the need
1. The maximum principal stress of the rock inclined
to examine the design rock stress values.
268 to the penstock side.
2. There is large deviation between two principal stresses
3. Prediction analysis and final design in a cross-section.
3. The stress concentrating zones of maximum principal
Analyses for predicting the behaviors of rocks during stress of over 25 MPa presented near the shoulder of
excavation were conducted using models that reproduced the tailrace side and at the corner of the penstock
the geological structures. Since highly-stiff triaxial com- side.
pression tests of boring cores showed that the rocks had
strain softening characteristics, the models were ana- The maximum principal stresses in these zones
lyzed by non-linear multi-step excavation FEM analysis increased as bench cut excavation advanced.
Changes in stress in the stress concentrating zones by
the installation of prestressed anchors (PS anchors),
which input stress, are schematically shown in Fig. 5.
An input of minimum principal stress by the installation
of PS anchors (confining force of DP) increased the
maximum supportable stress from s1 to s1qDs1,
showing that the installation is effective to restrict the
progress of the loosened zones.
In the actual design, PS anchors were more densely
installed in these stress concentrating zones than in the
other zones to actively restrict the progress of loosening
near the cavern. The final support arrangement in the
power plant cavern is shown in Fig. 6. Supports were

Table 1

Stratum Elastic modulus Peak strength Residual strength

GPa (vertical,
Cp (MPa) fp (8) Cr (MPa) fr (8)
horizontal)
M1 10 (14,7) 1.9 40 0.5 40
S1 22.5 (30, 15) 0.8 57 0.5 49
CG1 60 5.2 57 0.5 49
CG2 60 3.4 57 0.5 49
FS1 22.5 (30, 15) 3.4 57 0.5 49
FS2 30 3.4 57 0.5 49
Fig. 2. Geological cross-section.
 T. Maejima et al. / Tunnelling and Underground Space Technology 18 (2003) 223–232 225

Fig. 3. Initial rock stress.

cross-sectionally non-uniformly arranged in the cavern height of 3 m. The processes and stages of the power
to make rational use of the supports. plant cavern excavation are shown in Fig. 7.
The keyblocks that are generated in zones other than For each bench, enlargement excavation was con-
the loosened zones were decided to be individually dealt ducted using the ‘block partition method’. As shown in
with by observational construction techniques after mon- Fig. 8, excavation was conducted by first excavating the
itoring the actual behavior. center bench, then excavating the side benches and the
walls in a staggered manner, and finishing the side
4. Construction methods walls. The deformation of the peripheral rocks and the
progress of loosening was controlled by using the block
The upper arch-shaped part of the cavern was exca- partition method to be applied as a support before the
vated by the top heading and enlargement excavation next block to be excavated.
method. Enlargement excavation of the main cavern was The specifications of the supports are shotcrete (side
conducted using the bench cut method with a bench wall: 24 cm thick, roof and stress concentrating zones:

Fig. 4. Result of prediction analysis.

226 T. Maejima et al. / Tunnelling and Underground Space Technology 18 (2003) 223–232

Fig. 5. The mechanism for effect of input stress by PS anchor.

Fig. 6. Final support arrangement for the cavern.

 T. Maejima et al. / Tunnelling and Underground Space Technology 18 (2003) 223–232 227

as well as conventional strain and rock displacement

measurements. The items measured at the site are listed
in Table 2.
5.2. Observational design and construction procedures
Uncertain factors, such as initial stress that could not
be precisely monitored in the preliminary survey and
the effects of discontinuous surface along the tailrace
side, were not included in the standard design before
excavation but were decided to be individually designed
after confirming the actual behavior to make rational
use of the supports.
5.3. Program to optimize designs in the observational
construction technique
In order to take measures immediately after detecting
the behavior of these uncertain factors, prediction anal-
Fig. 7. Excavation process and stages of power station cavern. ysis was conducted to estimate the time when such
behavior was likely to be detected and the possible
32 cm thick), rock bolts (5–8 m long), and PS anchors reinforcement measures.
(10–20 m long). In order to determine the initial rock stress, it was
decided to monitor the absolute rock stress at the
5. Observational construction technique shoulders of cavern while spreading the ceiling.
Regarding the influence of the steep discontinuity in
5.1. Methods for evaluating the loosened zones layer of FS1 at the tailrace side, since the boundary of
FS1 and FS2 layer was not clear by the survey before
To monitor the redistribution of stress along the excavation, so, it predicted by setting up the model
progress of excavation, rock strain measurements and which considered the physical properties of the whole
acoustic emission (AE) measurements were conducted FS layer as an equivalent for the S1 layer in which the

Fig. 8. Excavation process for bench.

Table 2

Monitoring items Instruments Section

number
Daily management
Measurement of Multiple stage extensometer 10 sections
rock displacement Ls15;20 m (5, 6, 7, 8 stages)
Stress measurement Effective stressmeter 5 sections
of shotcrete
Load measurement Load cell (300 kN) 21 sections
of PS anchor
AE monitoring AE sensor 2 sections
and monitoring system
Stage management
Absolute rock stress Compact Conical-ended 2 sections
measurement Borehole Overcoring
(CCBO)
228 T. Maejima et al. / Tunnelling and Underground Space Technology 18 (2003) 223–232

discontinuity of steep inclination excels like the FS1 direction of the maximum principal stress action as the
layer. depth increased. The loosened zoned at the shoulders
The difference with the standard model appears in the reached depths of approximately 2 m.
development of the loosened zone at the time of the These results of behavior almost agreed with the
completion of 7th bench excavation (Fig. 9), and may results of a prediction analysis conducted using the
serve as the information for model selection. standard model, showing that both the direction and the
The time of the completion of 7th bench excavation intensity of the initial rock stress were valid. The
to which the loosened zone in side wall begins to supports during enlargement excavation were also decid-
progress was set as the object of step management. ed to be calculated using the standard model.
In addition, the support for middle part of cavern was
adopted as the support pattern determined by the anal- 6.2. Behavior and supports of the tailrace side during
ysis result using the low physical properties which can the excavation of the middle levels
correspond when it has the influence of discontinuities.
The behavior measurements of the rocks are shown
6. The result of observational construction in Fig. 10, which focuses on stress changes during the
procedures excavation of the seventh bench. Monitoring of the
excavation faces and borehole TV showed that there
6.1. Behavior and supports during enlargement exca- were three discontinuities, which were highly consecu-
vation of the cavern ceiling tive (L2, L3 and L4).
To confirm the progress of the loosened zones, abso-
The absolute rock stress values at the shoulders of lute rock stress was measured using the compact conical-
cavern are shown in Fig. 10, which were measured ended borehole overcoring method at the sixth and
using the compact conical-ended borehole overcoring seventh bench levels. The results of measurement
method. showed maximum principal stresses smaller than the
The tailrace side showed generation of concentrated initial rock stress up to the depth of 5 m and concen-
stresses that were larger than the initial rock stress at trated and uniaxial stresses in deeper than 7 m.
depths deeper than 2.3 m, and decreases in maximum The AE behaviors during excavation, which were
principal stress as the depth increased. monitored at the fourth bench, showed redistribution of
However, the penstock side showed a maximum stress concentration from the depth of 3 m during the
principal stress smaller than the initial rock stress at the excavation of the fifth bench to the depth of 7.5 m
depth of 1.9 m and counterclockwise changes in the during the excavation of the sixth bench. The depths

Fig. 9. Result of prediction analysis (loosened zone).

 T. Maejima et al. / Tunnelling and Underground Space Technology 18 (2003) 223–232 229

Fig. 10. Comparison of measurement result and prediction value.

where stress concentrated did not move thereafter. The the seventh bench was completed. This depth of the
AE frequency dropped at a depth of 6 m while excavat- loosened zone was approximately 2 m deeper than the
ing the sixth bench, but no such a drop was observed depth predicted using the standard model, which did not
at the depth of 7.5 m (Fig. 11). consider the effects of the discontinuous surfaces.
These results of observation showed that the loosened Regarding prediction analysis, since the distribution
zone was at the depth of 7.5 m when the excavation of of the discontinuity of high continuity became clear, the
230 T. Maejima et al. / Tunnelling and Underground Space Technology 18 (2003) 223–232

Fig. 11. Monitored AE event and frequency.

discontinuity was incorporated into the analysis model used in strain-softening analysis to predict behavior
and improvement in prediction accuracy was aimed for. (Fig. 10). The analyses could simulate the behavior at
Prediction analyses were conducted by modeling the the completion of the seventh bench excavation.
three highly consecutive discontinuities using anisotrop- A cause for larger loosened zones in this model than
ic parameters, which are the exfoliation and slippage of the standard model was likely to be the exfoliation and
the discontinuities, as proposed by Sakurai and Ine deformation of the discontinuous surfaces L2 and L3
(1986), Sakurai et al. (1998) and Sakurai et al. (in near the walls, which released the minimum principal
press). stress (along the radial direction of the cavern) and
These parameters and the elastic coefficients of the enhanced the uniaxial stress state in the rocks behind
discontinuities were then back analyzed (Fig. 12). The the cavern.
results of the inverse analysis were used to construct a The maximum loosened zones at the completion of
model that considered the discontinuity, which were excavation reached the depth of 10 m in the model that

Fig. 12. Discontinuity model for analysis.

 T. Maejima et al. / Tunnelling and Underground Space Technology 18 (2003) 223–232 231

Fig. 13. Distribution of rock displacement and strain.

reduced the entire physical properties of the FS layer, the tailrace side was estimated to be 7.5 m deep, which
but stopped at the depth of 7.5 m in the model that almost agreed with the value predicted using the model
reproduced the discontinuous surfaces. Therefore, the that considered the discontinuous surfaces (Fig. 10).
supports for levels lower than the eighth bench were These show that the prediction model that was revised
decided to be reduced (Fig. 6). after the completion of the seventh bench was effective
to accurately evaluate the effects of the discontinuities
6.3. Stability evaluation of the cavern at the tailrace side.

The displacement of the wall and the sectional strain 7. Conclusions

distribution are shown in Fig. 13 for a representative
cross-section at the completion of excavation. The strain This paper describes the observational construction
was discontinuously distributed along the depth direction technique for excavation of the underground cavern of
especially at the tailrace side. The sections of large the Kan-nagawa pumped storage power station.
strains almost agreed with the sections where the highly This large rock cavern was excavated by measuring
consecutive discontinuities existed, showing that the the stresses in the rock mass and the AE around the
behaviors of the surfaces were dominant. cavern which occurred due to the excavation, and using
To confirm the progress of loosened zones, absolute observational construction techniques especially for
stresses were measured along the measuring line of the monitoring loosened zones. The behavior at the final
fourth bench using the compact conical-ended borehole completion of excavation was correctly predicted by
overcoring method again (Fig. 10). The maximum monitoring the progress of the loosened zone along with
principal stresses at sections shallower than 6 m were the advancement of excavation and optimizing the ana-
smaller than the initial rock stress, but the stress con- lytical models during excavation.
centrated along the circumference of the cavern at depths This method was effective to optimize the support
deeper than 9 m. The loosened zone along the wall of system and conduct the excavation works safely. The
232 T. Maejima et al. / Tunnelling and Underground Space Technology 18 (2003) 223–232

progress of the loosened zones with discontinuity was cavern excavation. Rock Mechanics—A Challenge for Society,
well predicted by conducting strain-softening analysis Eurock 2001. pp. 443–448.
Maejima, T., Morioka, H., Mori, T., Aoki, K., 2001. Evaluation of
that considered the anisotropic deformation of the dis- the loosened zone on excavation of the large underground rock
continuities, which further improve the precision of the cavern. Modern Tunneling Science and Technology, IS-Kyoto 2001.
prediction analyses. pp. 1033–1038.
It is desirable that study continues on the methods for Sakurai, S., Ine, T., 1986. Strain analysis of jointed rock masses for
measuring, predicting and analyzing rock stresses to monitoring the stability of underground openings. Proceedings of
the International Symposium on Computer and Physical Modelling
make rational use of supports in large-scale underground
in Geotechnical Engineering, Bangkok. pp. 221–228.
caverns. Sakurai, S., Hiraoka, A., Hori, K., 1998. Strain-induced damage of
rocks. Proceedings of the 3rd International Conference on Mechan-
References ics of Jointed and Faulted Rock, Wien. pp. 21–27.
Sakurai, S., Akutagawa, S., Takeuchi, K., Shinji, M. (2003) Back
Maejima, T., Morioka, H., Mori, T., Aoki, K., 2001. Application of analysis for tunnel engineering as a modern observation method.
the observational construction technique for large underground Tunnelling Underground Space Technol, in press.