The effect of concrete expansion atOwen Falls power station, UgandaP. J. Mason, BSc, MPhil, PhD, CEng, FICE and J. D. Molyneux, BEng, CEng,MICE
& Cracking of concrete at Owen Falls power
station, Uganda, was diagnosed as being
due to concrete expansion caused by
alkali±aggregate reaction. Resulting struc-
tural movements had caused local over-
stressing and also de¯ections of installed
plant and equipment. The degree of
expansion varied according to the di�erent
cements used during construction. The
processes of detecting and clarifying the
implications of the various movements are
explained, together with measures taken
to mitigate immediate problems and
provide adequate monitoring to areas of
longer-term concern. Lessons are drawn
for the guidance of others investigating
similar phenomena.
Keywords: dams, barrages & reservoirs;
power stations (non-fossil fuel); rehabili-
tation, reclamation & renovation
IntroductionConstruction of the Owen Falls dam and power
station complex in Uganda started on site in
1951. The works comprised the damming of the
Victoria Nile a short distance downstream from
its source at Lake Victoria. A right ¯ank main
dam and sluice structure is separated by high
ground from a left ¯ank power station with ten
Kaplan turbines. The turbines were commis-
sioned in stages, the ®rst two in 1954 and the
last in 1968, giving a total installed capacity of
150 MW. The power station has provided the
overwhelming majority of the power in Uganda,
plus additional power for export, right up to the
present day. The location of the works is shown
in Fig. 1. A downstream view of the power
station is shown in Fig. 2 and an internal view
along the machine hall in Fig. 3. A chronology
of events and principal characteristics of the
schemes are given in Tables 1 and 2 respec-
tively.
2. In 1964 cracks were noticed in the con-
crete around the generator housing on machine
No. 4. Also in 1964, Lake Victoria, by then
e�ectively impounded by the dam and power
station, reached record levels with a rise of
more than 2 m above the levels which had been
carefully monitored since 1896. The Nile Waters
Agreement required additional releases to be
made through the sluices to re¯ect the level
rise, leading also, therefore, to higher tailwater
levels. The level rises increased both the
loading and the uplift on the power station and
other works. The fact that the cracks were ®rst
noticed in the same year that the rises occurred,
led to the inevitable suspicion that some form of
structural distress due to increased loading had
taken place. As this seemed, at the time, to be
limited to machine No. 4, no changes were made
to arrangements for the subsequent installation
of machines No. 9 and 10.
3. The power station and associated works
deteriorated rapidly during the 1970s as
Uganda passed through a period of political
instability, ®rstly through the Amin regime
(1971±79) and then through the Obote regime
(1980±85). The works were not inspected in
detail again by the original designers until
1983. By this time most turbines were operable
only at reduced load. This was largely due to
lack of spares and maintenance, although there
were also some machine misalignments and
clearance losses. The cracking in the power
station had also increased signi®cantly to
include a major crack up to 25 mm wide
through the machine hall ¯oor and running the
length of the power station (see Fig. 4). The
cracking at machine No. 4 had increased
signi®cantly and was mirrored at all other
machines. It should be noted that when the
inspections took place in 1983, general social
and security conditions in Uganda were still
quite dire with no local hotels available and
little infrastructure. Visiting engineers slept in
the power station.
4. Refurbishment works started in 1988 and
these included rewaterproo®ng the roof, making
underwater repairs throughout the works,
carrying out considerable amounts of stressed
anchoring in the power station to stabilize the
civil structure and generally refurbishing all
the mechanical and electrical equipment. As
part of this refurbishment the generators were
uprated from 15 to 18 MW, increasing the total
capacity of the station to 180 MW. Details of the
stressed anchoring which was carried out are
given elsewhere.1
5. During 1990, continued monitoring of the
cracks indicated that movement had not ceased
and that the underlying cause might be other
than simple structural overload. Indeed, the
crack monitoring results showed a broadly
linear trend of crack opening since 1973 which
did not, for example, seem to vary with changes
Peter J. Mason,
Director, Binnie,
Black & Veatch,
Redhill (formerly
Director, GIBB Ltd)
J. Dominic Molyneux,
Senior Project
Engineer, GIBB Ltd,
Reading
226
Proc. Instn Civ.
Engrs Wat., Marit.
& Energy, 1998,
130, Dec., 226±237
Paper 11726
Written discussion
closes 15 April 1999
in lake level. The present lead author became
involved at this stage and the following paper
broadly outlines the review and work that was
subsequently carried out to clarify and diag-
nose the cause of distress and to put in place
appropriate mitigating and monitoring mea-
sures.
Analysis of movements6. During the initial inspections in 1983,
samples of spalled concrete had been obtained
and were examined for potential distress such
as that caused by alkali±aggregate reaction
(AAR). This included analysis by thin section.
At that time no such distress could be detected.
The review in 1990 therefore focused on taking
a broad overview of what visible signs of
movement had occurred in order to visualize
overall patterns and see if this could shed
further light on underlying mechanisms.
7. It was noted that the patterns of cracking
and movement were very similar at most
machines, although focused more heavily on
machines No. 5 to 10. The pattern was therefore
initially viewed from a two-dimensional per-
spective as superimposed on a cross-section
through the power station, arbitrarily taken on
the centreline of machine No. 8 (see Fig. 5).
8. Monitoring the absolute and vector direc-
tions of crack movements in various parts of
the power station indicated that the down-
stream wall of the station was rotating down-
stream about a hinge point immediately above
the draft tube (see Fig. 6). It should be noted
that the power station was initially cast with
just the upstream and downstream walls as
®rst-stage concrete and with the latter heavily
reinforced to resist tailwater levels. This per-
mitted the machines to be erected and concreted
at a subsequent, second, stage.
9. It should also be noted that this down-
stream rotational movement was compatible
with two other observations. One was the main
longitudinal crack along the machine hall ¯oor.
This was up to 25 mm wide and, together with
other minor cracks, indicated a downstream
movement at that level of 32 mm. Secondly, the
overhead gantry crane rails were also known to
Mediterranean
Cairo
Egypt Red Sea
Sudan Khartoum
R. N
ile
KampalaOwenFalls
LakeVictoria
VictoriaNile
VictoriaNile
Tailrace
Headrace
Power station
Sluices
Main dam
Roadbridge
To Jinja
To Kampala
0 50 100 150
Scale: m
227
Fig. 1. Location plans
Fig. 2. Downstream
view of Owen Falls
power station
CONCRETE EXPANSION
AT OWEN POWER STATION
Fig. 3. View down the machine hall
Table 1. Owen Falls dam and hydropower complexÐa brief chronology
Date Event
1935 River Nile examined for hydroelectric potential
1947 Further survey of hydropower potential by Ugandan government
1948 Uganda Electricity Board formed
1949 Owen Falls hydroelectric complex planned
1951 First concrete placed at Owen Falls
1954 Machine No. 1 commissioned in January and No. 2 in April
Inauguration ceremony by HM the Queen on 29 April
1955 Machine No. 3 commissioned in January and No. 4 in August
1957 Machine No. 5 commissioned in January and No. 6 in February
50-year power agreement reached with Kenya
1958 Machine No. 7 commissioned in May
1959 Machine No. 8 commissioned in July
1964 Lake Victoria reaches unprecedented levels in May
First awareness of concrete cracking around machine No. 4
1966 Machine No. 9 commissioned in May
1968 Machine No. 10 commissioned in July
1971 Idi Amin seizes power in Uganda in a military coup
1973 Crack gauging commenced in power station by local sta�
1978 Targets installed to monitor downstream wall movements
1979 Idi Amin ousted by rebel forces
1980 Milton Obote returned to power
1983 Reinspection of power complex by team of UK engineers
1985 Milton Obote overthrown
1988 Start of refurbishment works on site
Table 2. Owen Falls dam and hydropower complexÐprincipal characteristics
Description Dimensions
Lake Victoria:
Catchment area 267 000 km2
Lake area 67 000 km2
Lake mean depth 40 m
Turbine generators:
Total number 10
Design head range 17´5±22 m
Originally installed output per machine 15 MW
Original ¯ow per turbine 96 m3/s
(now uprated to 18 MW with corresponding ¯ow
increase)
Sluices:
Total number 6
Size per sluice 3 m 6 5´1 m high
Design discharge per sluice 212 m3/s
Principal dimensions:
Dam crest road level 1136´15 m asl
Upstream max. water level 1135´00 m asl
Upstream min. storage level 1131´90 m asl
Max. tailwater level 1114´35 m asl
Min. tailwater level 1112´80 m asl
Nominal min. dam foundation level 1108´00 m asl
Nominal min. power station foundation level 1100´00 m asl
Length of gravity dam 726 m
Length of machine hall (excluding loading bay) 167´6 m
Width of machine hall 16´5 m
asl = above sea level.
Fig. 4. View of the longitudinal crack in the
machine hall ¯oor
228
MASON AND MOLYNEUX
be moving apart. Some years earlier, the over-
head crane had in fact jammed and crane
movement was reinstated by machining 9´5 mm
o� the bosses on one set of wheels to allow
them greater axle ¯oat.
10. It was also known that the power station
¯oor had risen in a number of locations. Again
these were concentrated between machines No.
6 and 10 but rises were as much as 76 mm (see
Fig. 7). Another indication of movement at
several locations was diagonal cracks in the
draft tube side walls (see Fig. 6).
11. After a reassessment of the evidence,
the 1990 review concluded that all these e�ects
could be broadly explained by an expansion of
the concrete around the machines. This would
exert a load on the downstream wall which was
one part of the structure relatively free to move.
It would also tend to cause cracking in the
original ®rst-stage concrete. Diagonal shear
cracking in the draft tube side walls would be
due to the force couple developed by the
expansive thrust downstream being resisted by
the draft tube foundations.
Initial modelling12. At this stage a simple two-dimensional
®nite element model was undertaken to assess
the e�ects of second-stage concrete expansion
around the turbines and their spiral casings
when viewed in the plan. The results are shown
in Fig. 8. It can be noted that for any single
machine, the concrete is unable to move later-
ally as it is restrained by neighbouring blocks,
nor can it move upstream. Movements are
therefore concentrated vertically and down-
stream. The downstream movement is ampli®ed
on the centrelines of the machines due to arch
action around the spiral casing. The model
therefore indicated the potential for the devel-
opment of voids downstream of the spirals and
also for vertical cracks between the ®rst- and
second-stage concrete in the stair wells between
the machines. Such vertical cracks were in fact
present on site. When the steel spiral casings
were subsequently drilled through downstream,
gaps between the steel and concrete of up to
15 mm were found.
13. The model also predicted that the mag-
ni®cation of upstream/downstream movement,
coupled with the lateral restraint against
expansion, would tend to produce ovality in the
water passages. This was also con®rmed by
clearance measurements around the tips of
the turbine blades, to the surrounding suction
cone.
Power station
Roof truss
Overhead crane
Draft tube deck
Generator
Lowerbracket
Columns
Tailrace
Spiralcasing
Turbinerunner
Suctioncone
Drafttube
Intake dam
229
Fig. 5. Typical cross-
section through the
power station
CONCRETE EXPANSION
AT OWEN POWER STATION
Different concretes14. A remaining concern was why di�erent
areas of the station had developed movements
in di�erent ways. This applied not only to the
power station proper but also to other parts of
the associated works such as the main dam and
intake structures.
15. In order to clarify this issue, the historic
records of concrete pours were examined for
various parts of the work. It was noted that
midway through construction the cement type
changed from imported Rugby cement from the
UK, to Tororo cement from Uganda's ®rst
cement factory which was commissioned in
1953. Fig. 9 shows patterns of level rise with
areas where Tororo cement was used. The
relationship between level rise and cement type
is clearly apparent.
16. An analysis of the Tororo cement indi-
cated that it had been produced from the local
volcanic material carbonatite which is very rich
in both alkalis and potash. This was certainly a
contributory factor to the accelerated e�ects of
alkali-aggregate reaction in those areas where
the local cement was used. Another factor was
that the coarse aggregate, though originally
classi®ed by the Uganda Geological Survey
under the broad generic name amphibolite, was
in fact amphibol-schist. This contains small
particles of reactive materials such as strained
quartz in a non-reactive matrix.
17. Typically, with this form of reaction,
symptoms are not seen in the early days after
construction. Gradually, alkali pore water seeps
from the cement paste into the large aggregate.
This a�ects any reactive particles, producing
silica gel and leading to internal splits and
expansion. The resultant local cracks even-
tually join, giving the appearance of structural
cracking, rather than random crazing. This
process is known as alkali±silicate reaction
(ASR) or `slow±late' reaction. It was discovered
much later than the more common AAR,
generally associated with reactive sand.2 The
e�ects of ASR, or slow±late reaction, are
typically not seen until ®ve to ten years after
construction.
18. The ®rst- and second-stage concretes in
the power station are described on record
drawings as 5/1� mixes. These in turn were
originally speci®ed as ratios of 5 cwt (254 kg) of
cement to 12 cu. ft (0´34 m3) of sand to 20 cu. ft
(0´57 m3) of coarse aggregate of maximum size
1� in. (38 mm). This is di�cult to interpret
precisely in modern terms without knowing
original bulking factors and densities. An
analysis of concrete core samples from the
power station, however, revealed cement con-
tents averaging 300 kg/m3 of concrete, plus or
minus about 50 kg/m3. Water/cement ratios
were back-analysed as between 0´52 and 0´60.
More importantly, the analyses revealed excep-
tionally high, equivalent alkali (sodium oxide)
contents of 2´5% of the cement and 7´5 kg/m3 of
the concrete.
19. The reinforcement used in the power
station concrete was mild steel throughout. The
average reinforcement density was 46 kg/m3.
Interestingly, the second-stage concrete around
the machines was completely unreinforced.
Years later this too was considered to be a
possible contributory factor towards the crack-
ing, although with hindsight, nominal reinforce-
ment would have done little to restrain the
expansive forces that were subsequently gener-
ated.
Power station stability20. Once the general e�ects of expansion
had been diagnosed, it was felt important that
the overall stability of the basic structural shell
be re-examined. This was done by a simple
structural model with superimposed move-
ments, and incorporated a standard roof truss
(see Fig. 10). It was clear that movements had
0 10 20
Crack vector scale: mm
Note: Arrows indicate the direction and magnitude of crack opening
Pronounced shear cracks,hairline to 2 mm
230
Fig. 6. Cross-section
through the power
station showing
downstream rotation
based on the vector
measurement of
cracks, also shear
cracking in the draft
tube side walls
MASON AND MOLYNEUX
generally started in the mid-1960s. This
allowed a rate of movement to be assumed of
approximately 1 mm per year horizontally at
machine hall ¯oor level. Overall stability and
structural adequacy were checked assuming a
further 30 years of expansion at the same rate.
21. Inherent stability of the overall struc-
ture was con®rmed but two local e�ects were
noted. One was that certain lower elements on
the roof trusses showed potential long-term
over-stress, depending on the potential yield of
structural connections on the upper wall. Sec-
ondly, it could be seen that the slender under-
water columns supporting the downstream deck
above the draft tubes were subject to rotation.
Diver examination had revealed that many of
these columns were severely distressed with
cracking, exposed reinforcement and consider-
able degradation of the concrete.
22. Measures to repair these columns were
discussed at some length and underwater trials
were carried out using epoxy repair techniques.
Dewatering by limpet co�erdams was also
considered. Eventually, however, the best way
forward emerged as a result of a focused, value
engineering study.
23. The value engineering study initially
considered various ways in which the down-
stream columns could be repaired but then,
alternatively, what other measures might be
available to give deck support. It was quickly
realized that drilling through the deck and
installing alternative circular steel columns
with structural connections to the tops of the
existing columns was a much more secure and
cost-e�ective way to proceed. Not only were the
repairs rapid but underwater diving time was
minimized. This was particularly important
given that the power station was in continual
use, with the client loath to cut back on
generation at any time unless absolutely essen-
tial. The repairs were successfully completed to
time and budget and at approximately half the
estimated cost of repairing the original columns
under water (see Fig. 11).
24. A visual inspection of the roof trusses
with regard to line, condition of paintwork,
rivets, etc. revealed no obvious signs of dis-
tress. With the evidence of expansion from
measurements of the crane rail gauge, and the
numerical modelling, this left the status of
the trusses uncertain. Therefore, before any
possible strengthening works were initiated,
extensometers were installed along the main,
lower truss elements to monitor movements and
provide evidence to justify any further expen-
diture.
25. Invar steel, bar extensometers, 5 m long,
were attached by clamping to the central lower
elements of the trusses 15 m above the machine
hall ¯oor. The instruments are read by trained
local monitoring sta� using a portable dial
gauge with the overhead gantry crane as a
mobile platform. Temperature is recorded along
with the dial gauge readings so that measure-
ments can be adjusted for the thermal expan-
Dow
nstr
eam
Ups
trea
m
Key section
E D F
C
B A
E
D
Fd(Fu shown dotted)
B
A
C
Machine
centreline
50 mm verticalmovement scale
Set
1 Set
2 Set
3 Set
4 Set
5 Set
6 Set
7 Set
8 Set
9 Set
10
Note: 1. Fu - Immediately upstream of main floor crack 2. Fd - Immediately downstream of main floor crack
231
Fig. 7. Isometric view
of machine hall ¯oor
rises
CONCRETE EXPANSION
AT OWEN POWER STATION
sion of the trusses. In fact, since the average
daytime temperature in Uganda varies little
during the year and the temperature in the
power house tends to be regulated even more,
corrections tend to be minor.
26. To date, results have been inconclusive
but suggest that, although movements may
occur, they could be subject to relief elsewhere.
Indeed, visual inspection did reveal some
cracking in the concrete columns supporting the
trusses. Another factor which may have helped
to relieve the build-up of stress in the trusses is
that the power station roo®ng was replaced
after the 1983 inspections. The original precast
concrete slab covering was replaced with a thin
synthetic membrane which has substantially
reduced the dead load carried by the trusses.
Turbine movement and three-dimensional modelling27. A remaining aspect, critical to the long-
term reliable operation of the station, was
understanding how historic and future expan-
sion might a�ect the turbines and generators. It
was noted, for example, that the rise in machine
hall ¯oor level had caused the turbine runners
to lift by as much as 76 mm compared to their
original elevation. A number of alternative
means to allow this situation to be redressed
were considered, including breaking out sup-
porting sole plates and generally lowering
them. There was, however, concern that this
might damage the structural integrity of the
concrete support.
28. It should be noted that in the same way
that the concrete was arching downstream in
plan between machines (see Fig. 8), it was also
arching vertically above each machine (see Fig.
12). This had led to horizontal cracking around
the turbine pits. This in turn meant that for
many years the generators had, in fact, been
sited on structural arches, or domes, founded
between the sets, rather than with the principal
load passing vertically down through the
turbine stay vanes. The fact that operation had
continued for most of the station's life in this
situation suggested that it was in fact accept-
able. Moreover, the situation is undoubtedly
reproduced in many of the other hydropower
stations in the world, similarly a�ected by
AAR. Attempts to grout such cracks would
have little e�ect, given that movements would
inevitably continue.
29. It was decided to produce a three-
dimensional model of a turbine block (see Fig.
13) and model the expansion using various
parameters for concrete and for restraint of
associated neighbouring concrete and steel
elements. This was done by varying the para-
meters and correlating them against measured
site movements on various concrete levels
around the turbines, notably those for the
stator, lower bracket and stay ring sole plates
(see Fig. 13). Detailed reference levels were
available for all three of these from the original
construction records. It was established that the
best correlations occurred with the Poisson's
ratio for the concrete, increased from 0´17 to
0´49, in e�ect allowing incompressible ¯uid,
rather than elastic, movement of the concrete.
Reductions in moduli were made elsewhere
where concrete was in tension and the struc-
tural restraint therefore provided by internal
steel reinforcement. The only machine restraint
found to be signi®cant was that provided by the
heavy, steel, lower bracket.
30. The model was essentially linear and
therefore somewhat of an approximation to the
inevitable non-linearity of the stress condition
actually occurring. It did, however, give a very
good indication of stress values, which were
broadly permissible, and also of patterns of
movement which replicated the downstream
rotation (see Fig. 14), machine hall ¯oor rises
and ovality of spiral casing and turbine pas-
sages, all predicted and measured earlier.
31. In order to restore machine operation
and to ensure future reliability, it was decided
Late
ral r
estr
aint
with
dow
nstr
eam
slid
ing
Fixed
Expanded profile
2D finite element expansion study
StairwellDownstream wall
Mass concrete
Differential expansiondownstream leadingto crack in stairwellbetween downstreamwall and mass concrete
Boundary of idealizedfinite element modelshown hatched
Upstream
Key plan on spiral
232
Fig. 8. Two-
dimensional ®nite
element model of
concrete expansion in
plan, on the centreline
of a turbine spiral
casing
MASON AND MOLYNEUX
to insert spacers in the main turbine/generator
shafts coupled with packing over the main sole
plates. The spacers were designed to accommo-
date future movement, over the next 25±30
years, with the compensatory packing under the
sole plates being progressively removed as
expansion continued. This is discussed in more
detail elsewhere3 and was developed jointly by
the civil and mechanical consultants.
32. Another result from this phase of the
work was establishing from surveys the actual
rises for the stator, lower bracket and stay ring
sole plates. It became clear from the values
obtained that the rises could be extrapolated
down to zero at approximately the base level of
the second-stage concrete around the machines.
This was a particularly useful ®nding as it
highlighted that the primary expansive force
was probably from the second-stage concrete
and was being exerted on the comparatively
less a�ected ®rst-stage concrete. The e�ect is
highlighted in Fig. 9 where the lower gallery,
founded on ®rst-stage concrete, has remained
close to design level whereas the upper gallery,
founded on second-stage concrete, has shown a
marked stepped rise from machines No. 6 to 10.
Stressed anchor monitoring33. In view of the likelihood of continued
concrete expansion, it was decided to use the
three-dimensional ®nite element model to
assess possible stress build-up in the stressed
anchors. Appropriate model node point separa-
tions were checked against the expansions
needed to calibrate the model against turbine
embedded part movements. The stress
increases that were predicted using this
approach were checked by carrying out lift-o�
tests on 24 representative anchors, out of a total
of approximately 170 installed.
34. The loads measured in the lift-o� tests
were generally lower than those predicted by
the ®nite element model. This result has been
found elsewhere in similar cases, although the
reasons are unclear. It may be related to strain
relief movements in the concrete immediately
under the end anchor plates, given the nature of
the AAR-a�ected concrete.
35. On six selected anchor heads, the cables
were completely de-stressed, permanent load
cells incorporated and the anchors re-stressed.
These will give a constant future record of
stress movements which can be used to assess
the likely condition of the anchors throughout
the station.
Future monitoring36. As described previously, monitoring at
Owen Falls began in 1973 with crack gauges.
Since then, as the potential problems were
recognized, a rather ad hoc monitoring regime
evolved. During 1994, with the newly acquired
understanding of movement causes, an e�ort
was made to rationalize all the existing instru-
mentation. A monitoring scheme was designed
to provide the information necessary to monitor
the concrete expansion and con®rm the safety
and serviceability of the dam and hydroelectric
power station. The scheme was implemented in
1994 as part of a contract to improve drainage
at the dam and also to investigate concrete
condition throughout the works.
37. A new network of thermally balanced,
double-skinned survey pillars and geodetic
survey targets was constructed. The geodetic
survey network is used to locate the power
station and draft tube deck to within an
accuracy of 2 mm. Such accuracy is deemed
su�cient to monitor drift movements caused by
the expanding concrete or any indication of
structural distress.
38. Survey points for a precise levelling
traverse through the power station were also
installed. These will enable monitoring of the
concrete and the machines and help evaluate
40
20
0
–20
20
0
–20
'Ris
e' in
mill
imet
ers
Set 10 Set 8
Set 9Set 7
Set 6
Set 5
Set 4
Set 3
Set 2
Set 1
Set 10
Set 8Set 9
Set 7 Set 6
Set 5
Set 4Set 3 Set 2
Set 1
Upper gallery
Lower gallery
2nd stage concretewith Tororo cement
2nd stage concretewith imported cement
Upper gallery1119·53
1117·09
1107·49
Stage 2
Stage 3
Stage 1
1115·26
1111·61
1109·17
Lowergallery
233
Fig. 9. Level rises in
relation to cement
type (elevations in m)
CONCRETE EXPANSION
AT OWEN POWER STATION
the need to change the sole plate packers
installed to accommodate this movement. As
the refurbishment is completed and the need for
expert expatriate sta� diminishes, it is intended
that all future monitoring work will be carried
out by local personnel. A comprehensive
manual was written describing the monitoring
requirements and outlining the history and
purpose of each of the 16 types of monitoring
required.
39. Monitoring generally is intended to
prevent events developing to undesirable levels
by giving warning in time for preventative
action to be taken. In order to serve this
purpose, monitoring data must be analysed and
interpreted promptly after instruments have
been read. It is no comfort to analyse data after
a disaster and be able to show that warning
signs were given; in fact, it may merely
demonstrate negligence.
40. At Owen Falls, two levels of alert have
been speci®ed for each type of monitoring. At
the (®rst) warning level, monitoring results
should be double-checked, frequency of read-
ings increased and a general increased level of
vigilance instigated. Depending on the circum-
stances, for instance if a number of independent
instruments demonstrate similar unusual
trends, it may be appropriate to notify the
power station management. At the (second)
alarm level, immediate action is required. All
instruments which could provide supporting
information should be rechecked and advice
should be taken on any action required to
safeguard the works.
41. The alert levels are assigned to pick up
short-term problems. Long-term trends of data
must be assessed annually and alert levels
adjusted accordingly. The actual values
assigned alter from instrument to instrument
depending on the usual scatter of results. Some
scatter is unavoidable due to seasonal ¯uctua-
tions, vibrations, accuracy of instruments,
human factors, etc. but they can be minimized
with careful procedures. The scatter of results
for any instrumentation scheme dictates the
minimum warning levels that can be set. Wider
bounds may be set if structural considerations
govern.
42. As a default case at Owen Falls,
warning levels were set at values more than
1´64 standard deviations above or below the
trend line in question. Assuming a normal
distribution, such values can be expected 10%
of the time. Alarm levels were set at 1´96
standard deviations above or below the trend
line in question, representing 5% of the time.
Supplementary criteria were set for extens-
ometer and crane gauge readings. Piezometer
readings were related to margins above tail-
water level. Drainage ¯ows were considered to
require judgmental assessment on an individual
basis.
43. For new instruments there is no scatter
of past readings to use to determine the
minimum alert levels. In order to evaluate the
expected scatter it was proposed that readings
were initially taken frequently. Several read-
ings in an hour would be necessary to establish
scatter due to vibration, hourly readings would
establish variation due to movements of the
sun, daily readings might be required to
measure variations due to reservoir level, and
monthly readings would be necessary to evalu-
ate seasonable changes. A knowledge of these
in¯uences is important in subsequently asses-
sing individual readings.
44. The long-term frequency of monitoring
must be a balance. Too frequent, and the
collected data will be overwhelming. Too sparce
and important movements could be missed. At
Owen Falls there are over 260 crack gauges.
These instruments provide the longest history
of movements in the complex but on the other
hand do not necessarily provide the best
measure of deformation in the short term
because of periodic stress build-up and sudden
relief. These instruments are therefore moni-
tored infrequently so that the history of move-
Deflected frame(exaggerated)
Original line ofpower station frame
Expandingconcrete
0 50 100
Deflection scale
234
Fig. 10. Structural
frame model of the
power station with a
32 mm horizontal
displacement
downstream at
machine hall ¯oor
level
MASON AND MOLYNEUX
ment is continued, but they are not used for
day-to-day monitoring. Measurements of the
overhead crane gauge combine all the crack
movements for any individual set and provide a
much easier way of assessing overall expansive
movements.
Conclusions and lessons learnt45. The development and diagnosis of AAR,
or ASR, at Owen Falls power station, followed
on remarkably parallel lines to the similar
diagnosis at Mactaquac power station in
Canada.4 In both cases, movements and crack-
ing were noticed. In both cases, initial conclu-
sions featured structural movements and
loadings or foundation movement. In both
cases, initial assessments of alkali-aggregate
reaction proved negative. In both cases, long-
term ASR or slow±late reaction proved to be the
case. The Mactaquac power station started
operation in 1968, problems were noticed in
1979, and the ASR eventually diagnosed in
1986.
46. In such cases, the ®rst signs of distress
are often maloperation of equipment and/or
some form of cracking. These cracks are then
monitored. The long-term trend of crack
development may be broadly linear; however,
it will often appear as a stepped pattern due
to normal measuring errors and due to the
build-up and release of pressure as silica gel
is formed. Furthermore, movement of one
crack may temporarily halt while the move-
ment is accommodated instead by an adjacent
crack.
Sole plateloads
Generator
Turbinepit crack
Probable(arching)load path
Suction cone
Draft tube
235
Fig. 11. Detail of a
typical bracket
connection between
old and new draft
tube deck support
columns
Fig. 12. Longitudinal
section through the
power station showing
structural arching
support to the
generators and
stators, and
horizontal cracking
around the turbine
pits
CONCRETE EXPANSION
AT OWEN POWER STATION
47. Apart from indicating whether the
movement is stable, accelerating or decelerat-
ing, the absolute monitoring of such cracks
does not always provide much useful informa-
tion. Inevitably, they are a secondary e�ect due
to expansion of concrete elsewhere. Of far more
immediate use and guidance is the overall
measurement of broad dimensions such as
height above foundation. In the case of Owen
Falls the ampli®ed width, as measured at the
crane rail, is probably the best indicator of
overall movement. The main gantry crane is
used as a stable, though temperature-depen-
dent, measuring bar and plates at either end are
®xed and used to measure o�sets to the rails. It
was also noted that one end of the power
station, at the loading bay beyond machine
No. 1, was completely uncracked and could be
taken as an original datum. Measurements of
crane rail separation on the machine centrelines
taken both proceeding down the station and
back and then averaged, gave a remarkably
good and consistent pattern of overall long-
term movements. These could be viewed
against global strains measured elsewhere, for
example overall dam height increases above
foundation.
48. It can be seen that the diagnosis of
alkali±aggregate reaction, eventually con®rmed
by concrete analysis, was accompanied by a
considerable amount of structural detective
work and intuitive analysis in which the skills
of the engineer were used to guide mathemat-
ical modelling rather than vice versa. This
must always be recommended as the way to
proceed.
49. Above all, the analyses were carried out
fully to understand the nature of the structure
both in its present state and how it would
continue to develop over the next 25±30 years,
which represents the economic life of the
installed turbines. They con®rmed that, with
regular monitoring, maintenance and adjust-
ment, the power station could continue to
operate e�ectively.
Acknowledgements50. The works described in this paper
were carried out between 1990 and 1997 when
the present lead author was a technical
director and subsequently director of GIBB
Ltd with responsibilities for the hands-on
direction of the contract at Owen Falls on
behalf of GIBB. The second author led
investigation work at site during 1994 includ-
ing coordination of the monitoring system and
rechecking stressed anchor loads. Particular
mention should be made of Mark Henning of
GIBB who so ably carried out the ®nite
element analyses; also of Peter Murray, the
GIBB inspector on site. Peter's career started
in Scotland with the manufacture of some of
the ®rst Owen Falls turbines. In the 1960s he
was based on site with the turbine Contractor
for the installation of later sets. Before
retiring in 1996 he spent eight years at Owen
Falls supervising their refurbishment. The
experience and knowledge of such men is
invaluable.
51. Mention should also be made of Dr Bill
French of Queen Mary and West®eld College
who so ably carried out the analysis of concrete
specimens taken from the site.
52. The work was carried out in conjunction
with Kennedy and Donkin Ltd of the UK who
were the lead consultants and particularly
responsible for the electro-mechanical aspects.
Richard Meileniewski of Kennedy and Donkin
Spiral
Suctioncone
Drafttube
Stator sole platesLower bracket sole plates
Stay ring
Fig. 13. Cut-away isometric view of the mesh for
the three-dimensional ®nite element model
Downstream
Suctioncone
Draft tube
Deflectedprofile
Originalprofile
Fig. 14. Section through the three-dimensional
®nite element model demonstrating downstream
rotation and ¯oor rise with second-stage
concrete expansion
236
MASON AND MOLYNEUX
carried out the detailed appraisal of turbine
movements from the electro-mechanical per-
spective. Electro-mechanical work in the power
station was funded principally by the UK
Department for International Development
(DFID), formerly the Overseas Development
Administration (ODA). Other works were
funded by the World Bank and the Common-
wealth Development Corporation. Lastly, the
authors would like to express their appreciation
to the Uganda Electricity Board, and in parti-
cular to Mr Alex Mugoya, for their cooperation
throughout the period of the authors' involve-
ment, and for their agreement to the publication
of this paper.
References1. ARCANGELLIRCANGELLI E. and STELLATELLA C. The use of pre-
stressed anchors at the Owen Falls refurbishment.
Water Power & Dam Construction, 1993, 45, 23±
30.
2. ICOLD. Alkali-aggregate Reaction in Concrete
DamsÐReview and Recommendation. International
Committee on Large Dams, Bulletin 79, Paris,
1991.
3. MUGOYAUGOYA A. Keeping hydro units aligned. Hydro
Review Worldwide, 1994, 2, No. 4, winter.
4. HAYWOODAYWOOD D. G., THOMPSONHOMPSON G. A., RIGBEYIGBEY S. J.
and STEELETEELE R. R. Engineering and construction
options for the management of slow/late alkali-
aggregate reactive concrete. International
Committee on Large Dams, Proceedings of the
16th Congress, San Francisco, 1968, Q62, R33,
575±588.
237
CONCRETE EXPANSION
AT OWEN POWER STATION