44
4 Historic Failures “Statistics”
Table 4.2 Recorded cause of failure, attributed causality scenario and related recorded failures
number (from USCOLD 1994, Fig. 4.4)
Recorded cause of failure
Attributed causality scenario
Recorded failures number
Slope instability, earthquake
and mine subsidence
Engineering error and
omission, excessive audacity
23 + 3 + 18 = 44
Over-topping
Poor management (mostly in
this example as a life time flaw
rather than an initial one)
17
Foundation
Poor investigation
9
Seepage and structural
Poor construction
10 + 7 = 17
Total = 87
of the e-IDC process, in order for the model to become amenable to analyses. Of
course, should detailed data on causality of failures become available in the future,
they could be readily included in the model and many assumptions made could be
released/replaced.
Table 4.2 shows that out of a total of 106 recorded failures (Fig. 4.4), 87 are from
known recorded causes (column 1), which were re-interpreted by us (column 2).
Failures World-Wide 1910–2009 Data
The data in Fig. 4.5 are a compilation (Azam and Li 2010) from UNEP, ICOLD,
WISE, USCOLD, USEPA.
Table 4.3 shows the same re-interpretation described above regarding the data of
ICOLD 1994 was performed for the 1910–2009 data. Out of 167 recorded failures,
145 are from known causes and 22 from unknown causes (column 1), which were
re-interpreted (column 2).
The Relative Split of Attributed Causality Scenarios for Initial Flaws
Using the data of Tables 4.2 and 4.3 it is possible to define a relative (%) split
of attributed causality scenarios stemming from project inception (Table 4.4). For
ICOLD 1994 (Table 4.2) over-topping had to be removed (as it was attributed to
management during lifetime), leaving us with 70 recorded failures with “known”
causes. For world-wide 1910–2009 data (Table 4.3) we eliminate poor management
for the same reason. Quake remains in the tally because if the dam fails under
seismic loading, then the design should be considered as faulty from the beginning
with respect to that loading.
We note a rather wide difference in the percentage split of causality, due to the poor
quality of the database, requiring the study to proceed with both values to include
uncertainties.
The Mitigations Models
In this study we consider two possible types of mitigations to be implemented during
the e-IDC development: M1, independent peer review; and M2, inspections. These
are described as follows:
4 Historic Failures “Statistics”
Table 4.2 Recorded cause of failure, attributed causality scenario and related recorded failures
number (from USCOLD 1994, Fig. 4.4)
Recorded cause of failure
Attributed causality scenario
Recorded failures number
Slope instability, earthquake
and mine subsidence
Engineering error and
omission, excessive audacity
23 + 3 + 18 = 44
Over-topping
Poor management (mostly in
this example as a life time flaw
rather than an initial one)
17
Foundation
Poor investigation
9
Seepage and structural
Poor construction
10 + 7 = 17
Total = 87
of the e-IDC process, in order for the model to become amenable to analyses. Of
course, should detailed data on causality of failures become available in the future,
they could be readily included in the model and many assumptions made could be
released/replaced.
Table 4.2 shows that out of a total of 106 recorded failures (Fig. 4.4), 87 are from
known recorded causes (column 1), which were re-interpreted by us (column 2).
Failures World-Wide 1910–2009 Data
The data in Fig. 4.5 are a compilation (Azam and Li 2010) from UNEP, ICOLD,
WISE, USCOLD, USEPA.
Table 4.3 shows the same re-interpretation described above regarding the data of
ICOLD 1994 was performed for the 1910–2009 data. Out of 167 recorded failures,
145 are from known causes and 22 from unknown causes (column 1), which were
re-interpreted (column 2).
The Relative Split of Attributed Causality Scenarios for Initial Flaws
Using the data of Tables 4.2 and 4.3 it is possible to define a relative (%) split
of attributed causality scenarios stemming from project inception (Table 4.4). For
ICOLD 1994 (Table 4.2) over-topping had to be removed (as it was attributed to
management during lifetime), leaving us with 70 recorded failures with “known”
causes. For world-wide 1910–2009 data (Table 4.3) we eliminate poor management
for the same reason. Quake remains in the tally because if the dam fails under
seismic loading, then the design should be considered as faulty from the beginning
with respect to that loading.
We note a rather wide difference in the percentage split of causality, due to the poor
quality of the database, requiring the study to proceed with both values to include
uncertainties.
The Mitigations Models
In this study we consider two possible types of mitigations to be implemented during
the e-IDC development: M1, independent peer review; and M2, inspections. These
are described as follows: