Professional Documents
Culture Documents
A
number of earth dams have failed or suffered
large displacements during past earthquakes. captures the element data observed in cyclic laborato-
Fig. 1 shows that earthquake is a major cause ry tests and verified by comparison with model tests
(the second, accounting for 17 per cent) of failures of and field experience is applied to the Mochikoshi dam.
tailings dams [USCOLD, 19941]. Based on such analyses, implications for the design of
In most cases, the damage has occurred as a result of impoundment structures to resist seismic loading are
a large reduction in the stiffness and strength of soil examined.
(liquefaction). Classic examples of liquefaction dam-
age are the behaviour of the San Fernando dams dur- 1. Soil liquefaction
ing the 1971 San Fernando earthquake, California. Seismic liquefaction refers to a sudden loss in stiffness
The crest of the upper dam moved downstream about and strength of soil as a result of cyclic loading effects
1.5 m, while a flow slide on the upstream side of the of an earthquake. This loss arises from a tendency for
lower dam, some moments after the severe shaking, soil to contract under cyclic loading, and if such con-
removed the crest of the lower dam as shown in the traction is prevented or curtailed by the presence of
photograph (from:www.ce.washington.edu). water in the pores that cannot escape, it leads to a rise
A number of mine tailings impoundment dams have in porewater pressure and a resulting decrease in
also suffered severe damage as a result of liquefaction effective stress. If the effective stress drops to zero
during past earthquakes. The most notable examples (100 per cent porewater pressure rise), the strength
are the Mochikoshi dams in Japan, which failed dur- and stiffness also drop to zero and the soil behaves as
ing the 1978 Izu-Ohshim-Kinkai earthquake as a a heavy liquid. However, unless the soil is very loose,
result of liquefaction-induced flow slides, resulting in it will dilate and regain some stiffness and strength, as
release of the tailings, as shown in Fig. 2. One dam it strains. If this strength is sufficient, it will prevent a
failed during the shaking, while a second failed 24 flow slide from occurring, but may still result in
hours later. excessive displacements commonly referred to as lat-
Our understanding of the seismic behaviour of earth eral spreading. In addition, even for level ground con-
structures has increased considerably in the past 30 ditions where there is no possibility of a flow slide and
years as a result of: lateral movements may be tolerable, very significant
• observations from field case histories; settlements may occur because of dissipation of
• extensive laboratory testing of soil elements under excess porewater pressures during and after the period
cyclic loading; of strong ground shaking.
Fig. 1. Statistics on
the causes of • model testing of earth structures under simulated
failure of tailings earthquake loading; and, 1.1 Assessment of liquefaction
dams. • development of numerical modelling procedures. Liquefaction assessment involves addressing the fol-
lowing concerns:
• Will liquefaction be triggered in significant zones of
the soil structure for the design earthquake? and,
• If so, could a flow slide occur? and if not,
• Are the displacements tolerable?
technical drillings conducted about three weeks after and ϕ’ = 35º, and the permeability was reported to be
the failure was about zero to 2 above a depth of 15 m 10-4 cm/s. The average N value of starter dam was
and about 3 to 7 at greater depths. Ishihara [198416) reported to be about 5. The above properties were
noted that this low N value might have been caused by reported by Ishihara [198416]. This material was con-
soil remolding, as a result of liquefaction during the sidered to be non-liquefiable.
shaking. As demonstrated by Beaty and Byrne [19978],
for the flow failure at the Mufulira Mine in Zambia in 2.3 Numerical modelling
1970, mixing of sandy silt soils significantly reduced Dynamic analysis of the Mochikoshi Dam No.1 was
its strength, leading to further strength and stiffness carried out using the UBCSAND constitutive model
loss after liquefaction had been triggered. It is likely for the tailings materials. This is based on the elastic-
therefore that the original blow counts of the tailings at plastic stress strain model proposed by Byrne et al.
Mochikoshi were somewhat higher. It was assumed [199524], and has been further developed and extended
that the average N value was 2. The equivalent by Puebla et al. [199725; Puebla, 199926], and Beaty and
(N1 )60-cs corrected for overburden pressure and fines Byrne [199827]. The model has been successfully used
content based on Seed [198521] and Youd et al. [200122] to analyse foundation liquefaction and also to predict
was estimated to be 6 and 11 in the upper and lower the behaviour of remediation measures comparing cen-
parts of the slime, respectively. trifuge testing results [Seid-Karbasi, 200320].
The results of static triaxial tests on undisturbed tail- It is an incremental elastic-plastic model in which
ings indicated zero cohesion and a friction angle ϕ’ the yield loci are lines of constant stress ratio. The
varying between 30 and 39 degrees. The liquid limit of flow rule relating the plastic strain increment direc-
tailings varied between 27 and 31 per cent, whereas the tions is non-associated and leads to a plastic potential
moisture content ranged from 36 to 37 per cent. A water defined in terms of dilation angle. The model is
content greater than the liquid limit indicates that the implemented in the commercial computer code FLAC
tailings would be very susceptible to strength loss as (Fast Lagrangian analysis of Continua, ITASCA,
they sheared, as also noted by Bray et al [200423]. The 200028].
permeability of the materials was 10-4 and 10-7 cm/s in The appropriate parameters for the model can be
horizontal and vertical directions, respectively. The obtained directly from cyclic testing of undisturbed
void ratio of the tailings was 0.98 and 1.0 and specific samples from the site, or indirectly from field experi-
gravity was 2.72 and 2.74, respectively, for the sandy ence with similar soils during past earthquakes. The
silt and silt portions of the tailings [Ishihara, 198416]. common practice is the indirect approach, with lique-
The starter dam itself was placed and compacted by faction response expressed in terms of penetration
bulldozers, and consisted of a mixture of weathered resistance, and this approach was used here. The
tuffs and volcanic ash obtained from the borrow pit UBCSAND model has been calibrated to reproduce
adjacent to the dam. It comprised a mix of gravel, sand the Youd et al. [200122] triggering chart, which in turn
and silt and had a 65 per cent fines content. Its unit is based on field experience during past earthquakes
weight ranged between 14 and 19 kN/m3. The natural and is expressed in terms of Standard Penetration Test
moisture content ranged between 30 per cent and 60 resistance value, N1(60). The model properties to obtain
per cent, and the void ratio was 1.1 to 2.6. Triaxial such agreement are therefore expressed in terms of
tests on undisturbed samples resulted in C’=25 kPa N1(60). It has also been calibrated with cyclic simple
shear test data for Nevada sand as well as Fraser river
sand, and predicts triggering as well as post-triggering
response in close agreement with the data, as shown in
Fig. 4. The agreement is obtained by selecting an
N1(60) to give the best fit to data. In this way an N1(60)
value equivalent to the known relative density, Dr, of
the laboratory test sample is obtained.
The model grid, together with material types repre-
senting different parts of the dam in the analysis, is
illustrated in Fig. 5.
In the first-stage static analysis, the Mohr-Coulomb
model with stress-dependent materials properties was
used for all the parts. The soil was treated as equivalent
elastic and isotropic, using secant shear (G) and bulk
(B) moduli which vary with the stress level as follows:
G = kg . Pa . (σ’m /Pa)n ... (1)
B = kb . Pa . (σ’m /Pa) m
... (2)
Fig. 4. Comparison of
predicted and
measured response
for Fraser river sand
(a) stress path (b)
excess pore pressure
ratio (c) number of
cycles, test data from
Sriskandakumar,
[200331]. Fig. 5. Dam model grid, comprising various materials.
Soil kg n kb m ϕ C γt Permeability
(kPa) (kN/m3) (cm/s)
3. Lessons learned
Lessons learned from field experience, laboratory
tests, and analyses are:
• The seismic failure of Dam No.1 at the Mochikoshi
impoundment probably occurred as a result of liquefac-
tion of the tailings. The presence of layers of plastic silt
having a natural water content in excess of the liquid limit,
probably caused a further strength loss resulting in flow of
the tailings once significant displacements occurred.
• A contributing factor to the failure may have been the
presence of low permeability barrier layers which impede
vertical drainage of excess porewater pressure. This can
greatly reduce stability as they may cause a water bubble
to form at the base of the layer during or after shaking, as
observed by Kokusho [200030] and as shown in Fig. 14.
• Such barrier layers probably caused the delayed
failures at the Lower San Fernando dam and at the
Mochikoshi No. 2 dyke.
References:
1. US Committee on Tailings Dams, “Tailings dams
incidents”, 1994. (Quoted by Lo and Klohn, ICOLD
Symposium on Seismic and Environmental Aspects of Dam
Design, Santiago, Chile; 1996.
2. Newmark, N. M., “Effects of Earthquakes on Dams and
Embankments”, Geotechnique, Vol. 15, No. 2, 1965.
3. Arulanandan, K., and Scott, R.F., “Verification of
numerical procedures for the analysis of soil liquefaction
problems”, Proceedings, International Conference on the
Verification of Numerical Procedures for the Analysis of Soil
Liquefaction Problems, Vols. 1 and 2, Balkema, Rotterdam,
The Netherlands; 1993.
4. Dafalias, Y.F., “Bounding surface plasticity. I: mathematical
Fig. 14. Water bubble under the barrier layer caused by shaking. foundation and the concept of hypoplasticity”, Journal of
Engineering Mechanics, ASCE, 112 (9), pp. 966-987; 1986.
• The bubble effect can be prevented by vertical 5. Prevost, J.H., “DYNA1D: A computer program for non-
drains that penetrate the barrier layers, as illustrated in linear site response analysis”, Technical Report NCEER-89-
Fig. 15. The design of these drains can be assessed 0025, National Center for Earthquake Engineering Research,
from dynamic coupled flow-effective stress analyses. SUNY at Buffalo, New York, USA; 1989.
• The Mochikoshi dam could have been stabilized by 6. Zienkiewicz, O.C., Chan, A.H.C., Pastor, M., Paul, D.K.,
a free-draining buttress fill or by a stabilized column and Shiomi, T., “Static and dynamic behavior of soils: a
rational approach to quantitative solutions. Part I: fully
as shown in Figs. 16a and b. saturated problems”, Proceedings, Research Society
• The dimensions and location of remediation mea- London, A429; 1990.
sures can be optimized by dynamic analyses. 7. Byrne, P.M., Roy, D., Campanella, R.G., and Hughes, J.,
Stabilization can be achieved by densification, drainage “Predicting liquefaction response of granular soils from
or bonding of soil particles to prevent liquefaction. pressuremeter tests”, ASCE National Convention, San
Diego, USA; ASCE Geotechnical Special Publication 56;
October 1995.
4. Conclusion 8. Beaty, M. H. and Byrne, P. M., 1997. “Post-liquefaction
A number of impoundment dams have failed during past shear strength of granular soils: theoritical/conceptual
earthquakes as result of soil liquefaction. Plastic sandy issues”, Proceedings, Workshop on Post-Liquefaction Shear
Shear Strength of Granular Soils, Urbana-Champion,
silt layers may have water contents greater than their liq- Illinois, USA; April 1997.
uid limit, in which case they may lose more strength 9. Elgamal, A.-W., Parra, E., Yang, Z., Dobry, R., and
when significant displacements are induced by seismic Zeghal M. 1999. “Liquefaction constitutive mode”,
loading. The failure of Dam No. 1 at the Mochikoshi Proceedings, Physics and Mechanics of Soil Liquefaction
impoundment probably occurred in this way. Symposium, Balkema, Rotterdam, The Netherlands; 1999.
Laboratory testing suggests that clean loose sands are 10. Kramer, S., and Arduino, P., “Constitutive modeling of
unlikely to suffer a flow slide, because, although they cyclic mobility and implications for site response”,
Proceedings, 2nd International Conference on Earthquake
can be triggered to liquefy, their undrained strengths are Geotechnical Engineering, Balkema, Rotterdam, The
generally adequate for stability, unless they are very Netherlands; 1999.
loose. However, if the sands contain low permeability 11. Marcuson, W.F., III, Ballard, R.F., Jr., and Ledbetter,
silt layers which impede drainage, water bubbles and R.H., “Liquefaction failure of tailings dams resulting from
complete loss of strength can occur. The delayed failure the Near Izu Oshima earthquake, 14 and 15 January, 1978”,
Proceedings, 6th Pan-American Conference on Soil
of dyke No. 2 at the Mochikoshi impoundment as well Mechanics and Foundation Engineering, Lima, Peru, Vol. 2;
as the delayed failure of the lower San Fernando dam 1979.
may well have occurred in this manner. ◊ 12. Marcuson, W.F., III., “Visit to Japan to observe damage
which occurred during the Near Izu Oshima earthquakes
January 14 and 15, 1978”, Miscellaneous Paper GL-79-20,
US Army Corps of Engineers Waterways Experiment
Station, Vicksburg, Missouri, USA; 1979.
13. Okusa, S. and Anma, S. “Slope failures and tailings dam
damage in the 1978 Izu- Ohshima-Kinkai earthquake”,
Journal of Engineering Geology, No. 16; 1980.
14. Okusa, S., Anma, S., and Maikuma, H., “Liquefaction of mine
tailings in the 1978 Izu- Ohshima-Kinkai earthquake, central
Japan”, Proceedings, Vol. 3, 7th World Conference on
Earthquake Engineering, Istanbul, Turkey; September 1980.
Fig. 15. Drains to curtail the bubble effects. 15. Okusa, S., Anma, S., and Maikuma, M., “The propagation of
liquefaction pressure and delayed failure of a tailings dam dike
in the 1978 Izu-Oshima-Kinkai earthquake”, Proceedings, 8th
World Conference on Earthquake Engineering, July 21-28, San
Francisco, CA, USA; Vol. 1, 1984.
16. Ishihara, K., “Post-earthquake failure of a tailings dam due
to liquefaction of the pond deposit”, Proceedings, Vol. 3,
International Conference on Case Histories in Geotechnical
Engineering, Rolla, Missouri, USA; May 1984.
17. Ishihara, K., Yasuda, S., and Yoshida, Y., “Liquefaction-
induced flow failure of embankments and residual strength of silty
sands”, Journal of Soils and Foundations, Vol. 30, No. 3, 1990.
18. Jitno, H. and Byrne, P.M., “Predicted and observed liquefaction
response of Mochikoshi tailings dam”, Vol. 2, Proceedings, First
International Conference on Earthquake Geotechnical
Engineering, Tokyo, Japan; 14-16 November, 1995.
19. Olson, S. M., “Liquefaction analysis of level and sloping ground
using field case histories and penetration resistance”, PhD
Fig. 16. Remediation of Mochikoshi tailings dam: (a) buttress; Thesis, University of Illinois, Urbana-Champaign, USA; 2001.
(b) stabilizing column. 20. Seid-Karbasi, M., “A numerical study on liquefaction