A back-analysis of the Beliche Dam in southern Portugal is described.
This dam was the subject of a 'Class A' prediction which grossly under
predicted the actual settlement at the end of construction. The analys
is described here, which incorporated an elasto-plastic critical model
applied in terms of effective simulated the effect of collapse settle
ment and by so doing reproduced a maximum settlement which was close t
o that measured. The earlier underprediction is attributed mainly to t
he fact that that analysis did not include the collapse settlement due
to the partial impounding of the reservoir which occurred before cons
truction was complete. The back-analysis described here covers the per
iod from the start of construction in 1984 to approximately four years
after the first complete impounding in January 1988. Comparisons are
made between the calculated and measured movements and stresses. Apart
from the good agreement with the maximum recorded settlement, the mea
sured settlements generally exceeded the calculated values. Part of th
e underprediction of settlements is attributed to creep which the mode
lling did not attempt to simulate and part to the in situ relative den
sity being less than that used in the tests from which the parameters
were determined. The computed stress distributions indicated a favoura
ble load transfer from shell to core, thus confirming the safety of th
e dam against internal erosion due to hydraulic fracture. It was, howe
ver, possible to estimate from the settlement records how much of the
settlement is due to creep, and this has been done. The conclusions un
derline the usefulness of the collapse settlement methodology.