Sei sulla pagina 1di 13

See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/288595356

Health monitoring of concrete dams: A


literature review

Article in Journal of Civil Structural Health Monitoring October 2014


DOI: 10.1007/s13349-014-0079-2

CITATIONS READS

8 393

4 authors, including:

Patrick Bukenya Pilate Moyo


Cape Peninsula University of Technology University of Cape Town
9 PUBLICATIONS 9 CITATIONS 70 PUBLICATIONS 774 CITATIONS

SEE PROFILE SEE PROFILE

Hans Beushausen
University of Cape Town
91 PUBLICATIONS 518 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Design of Sustainable Concrete Structures View project

structural health monitoring of dams View project

All content following this page was uploaded by Patrick Bukenya on 14 April 2016.

The user has requested enhancement of the downloaded file.


Health monitoring of concrete dams: a
literature review

P.Bukenya, P.Moyo, H.Beushausen &


C.Oosthuizen

Journal of Civil Structural Health


Monitoring

ISSN 2190-5452
Volume 4
Number 4

J Civil Struct Health Monit (2014)


4:235-244
DOI 10.1007/s13349-014-0079-2

1 23
Your article is protected by copyright and
all rights are held exclusively by Springer-
Verlag Berlin Heidelberg. This e-offprint is
for personal use only and shall not be self-
archived in electronic repositories. If you wish
to self-archive your article, please use the
accepted manuscript version for posting on
your own website. You may further deposit
the accepted manuscript version in any
repository, provided it is only made publicly
available 12 months after official publication
or later and provided acknowledgement is
given to the original source of publication
and a link is inserted to the published article
on Springer's website. The link must be
accompanied by the following text: "The final
publication is available at link.springer.com.

1 23
Author's personal copy
J Civil Struct Health Monit (2014) 4:235244
DOI 10.1007/s13349-014-0079-2

ORIGINAL PAPER

Health monitoring of concrete dams: a literature review


P. Bukenya P. Moyo H. Beushausen

C. Oosthuizen

Received: 5 June 2013 / Revised: 1 April 2014 / Accepted: 15 April 2014 / Published online: 11 May 2014
Springer-Verlag Berlin Heidelberg 2014

Abstract Health monitoring of concrete dams has Lake dam that claimed 236 people [1] and Austin dam
become a topic of great importance and involves moni- failure which claimed lives of 78 people [2]. These disas-
toring the static and dynamic behavior of large dams. This ters and their consequences can be minimized if appro-
paper provides a literature review on the subject of health priate dam safety monitoring and surveillance strategies are
monitoring of concrete dams based on static and dynamic implemented.
tests. The intent of this paper is to evaluate case studies in Health monitoring of concrete dams commonly referred
health monitoring of concrete dams whereby the reader can to as dam surveillance is a process of assessing the struc-
reflect on the various techniques that have been used in tural integrity of dams through detection of any abnor-
analyzing data in this field. The knowledge obtained will mality as early as possible which could be an indicator of
be used to highlight the state of the art and consequently some danger and allowing sufficient forewarning for the
advance the research in this field of dam engineering. implementation of appropriate corrective measures. Regu-
lations and guidelines have been developed to help dam
Keywords Static monitoring  Dynamic monitoring  engineers to monitor the structural integrity of dams.
Ambient vibration tests  Static properties  Dynamic Examples of regulations and guidelines include the UK
properties Reservoir which gives dam owners the responsibility of
monitoring dams [3], Australian National Commission on
Large Dams [4], South African National Commission on
1 Introduction Large Dams [5] and in more detail with respect to for-
malized monitoring by the International Commission on
Concrete dams support economies by providing water for Large Dams [6].
irrigation, hydroelectricity generation and flood control. The traditional procedure for evaluating the structural
The performance of these structures under operational and integrity of dams is through visual inspections. Visual
environmental loads may decrease over time due to age- inspections are conducted by experienced engineers who
related deterioration, floods and other factors. If concrete later recommend actions to be carried out on dams under
dams are not well managed and maintained, failure may inspections to solve the identified problems. Visual inspec-
occur leading to economic and life losses. Examples of tions have various shortcomings such as high manpower
disasters that have occurred in the world include Canyon demand, often insufficient frequency and inaccessibility of
critical parts of the structure under surveillance. For exam-
ple, if damage develops on the upstream side, under water
P. Bukenya (&)  P. Moyo  H. Beushausen
Department of Civl Engineering, University of Cape Town, level or contact between dam and foundation rock, it is dif-
Cape Town, South Africa ficult to observe them directly. The consequential shortage of
e-mail: bknpat002@myuct.ac.za information pertaining to the structural condition may result
in an erroneous condition evaluation and hence uninformed
C. Oosthuizen
Department of Water Affairs South Africa, decisions regarding the maintenance of the structure. The
Pretoria, South Africa shortcomings of visual inspections can be overcome by the

123
Author's personal copy
236 J Civil Struct Health Monit (2014) 4:235244

use of full-scale field tests. Field tests are conducted to opening and closing of joints, crack opening, displace-
determine and monitor static and dynamic properties of the ments and strains which are measured accurately by
structure under environmental conditions. Static properties instruments. A lot of data are collected from the different
and dynamic properties are determined through static and instruments; hence interpretation of these data is important
dynamic monitoring processes, respectively. in structural health monitoring (SHM) of concrete dams. In
In static monitoring, environmental characteristics such static monitoring, data interpretation enables in taking into
as wind speed, air temperature and water temperature are account of any effects of loads separately on the structural
measured as the deformation of the dams, strains and stres- behavior of dams. This is done either using statistical and/
ses. Data obtained from static monitoring is either analysed or deterministic models. Statistical models establish rela-
using statistical and or deterministic models [7]. Statistical tionships between the present and the past behavior of the
models are based on correlations between environmental dam. The accuracy of this model depends on the amount
factors (reservoir water level, ambient temperatures, wind) and reliability of the available data. This approach pro-
and dam responses (displacements, pressures, stresses). vides an answer to the question: Does current dam
These correlations are estimated by performing statistical behavior correspond to behavior observed in the past?
analysis of historical data [8]. In the dynamic monitoring Deterministic models establish the relationship between
approach, structural responses in form of accelerations due to the loads and the dam response defined by a structural
vibrations are measured. Accelerations are obtained using analysis. This approach answers the question: Does current
either forced vibration tests or ambient vibration tests [9]. dam behavior correspond what it should be? Such
Dynamic properties (natural frequencies, mode shapes and approaches try to correlate field results and analytical
damping ratios) can then be extracted from the accelerations results from a model.
using modal analysis techniques [10]. Bianchi and Bremen [7] presented procedures to carry-
Significant research has been done in the area of health out investigations on the long-term behavior of Ferden and
monitoring of dams with the dual purpose of understanding Roggiasca dams in Switzerland. The purpose of this study
the behavior of these structures and prevention of failure. was to detect any changes in the dam behavior both before
Therefore, it is better practice that dam engineers and the snowfall and after the break period. During the 5 week
researchers have knowledge on which methods have been long period the temperature within the concrete and res-
used in the analysis of data collected from static and dynamic ervoir level had varied so that a careful evaluation of the
monitoring of dams. This paper provides a literature review exceptional loading conditions was needed. A deterministic
of the case studies in health monitoring of dams whereby the tool called MIC based on probabilistic and deterministic
reader can reflect on the methods that have been used and models was developed to carry out the evaluation of these
understand advantages and limitations of the various two dams. This too was used for continuous comparison of
approaches. The knowledge obtained will be used to high- the measured dam behavior to expected behavior of these
light the state of the art in health monitoring of concrete dams structures. Probabilistic model was used for monitoring
and to point out future research needs. The paper is organized long-term behavior of dam while deterministic models
as follows: Sect. 2 presents static monitoring of dams were based on measured-expected using a structural model
addressing the name of the dam, how long were the data and which considered static and thermal effects. Results show
which methods were used in the analysis. Tremendous that crest dam displacements of 11 mm had occurred after
amount of literature concerning dynamic monitoring of the snow fall on Ferden dam. To assess the long-term
concrete dams have been published over years; however, behavior of the Roggiasca dam accurately, MIC was used
only representative papers and case studies regarding to plot the development of reference deformation over a
ambient vibration monitoring of dams are reviewed in Sect. long period of 20 years. The authors concluded that the
3. Case studies related to seismic monitoring of dams are not advanced deterministic approach led to reliable explana-
reviewed as they are out of scope of this paper. Section 4 tions of and solutions to extraordinary situations.
gives a general summary of the literature review provided in Behrouz [8] analyzed strain data collected for
the paper highlighting the progress that has been made in 17 months from Idukki dam in India. The hydrostatic-
health monitoring of dams. Finally, Sect. 5 identifies the season-time (HST) model was used to model and predict
critical issues in the review and concludes the paper. the future strains of the dam. Multiple linear regression
analysis was used to determine the dependency of the
maximum strains with the reservoir level, time and ambient
2 Static monitoring of dams temperature. Principal component analysis was used to
detect any outliers hence identifying any anomalies in the
Static monitoring of dams involves measurement of static reference model and for subsequent comparison of future
factors such as ambient temperatures, reservoir level, values to the model. The author concluded that the HST

123
Author's personal copy
J Civil Struct Health Monit (2014) 4:235244 237

model explained a large portion of variances and is accu- it provided a reasonable separation of various factors of
rate in predicting strains. influence.
Pytharouli and Stiros [11] analyzed geodetic monitoring Mata [15] compared two statistical models namely
data for the long term behavior of Landon dam in Western multiple linear regression (MLR) and neural network (NN)
Greece. Horizontal and vertical displacements were mea- models for characterization of Alto Rabago dam behavior
sured for a period of more than 30 years used in the ana- under environmental loads of reservoir level and external
lysis of Landon dam. Discrete Fourier transform and Lomb temperature of structural responses (crest displacements).
normalized period gram methods were used to investigate Displacements at any point of the dam were strongly
if both dam deformation and reservoir level fluctuations related to the corresponding variation in water level in the
corresponded to periodic functions of the same period. reservoir. The observations were used for the computation
Results showed that for periods greater than 30 years long, of the MLR and NN models. MLR and NN models were
during which floods and earthquakes occurred, Ladon dam generated and calibrated on the basis of experimental data
retained their structural integrity and their deformation was time histories (over 25 years) of reservoir level and
kept at low levels. external temperatures and of crest displacements. The NN
Chouinard et al. [12] demonstrated the benefits of model showed flexibility and proved to be more adequate
multivariate statistical analysis to study the long term for months with extreme temperatures than the MLR
behavior of Daniel Johnson dam. The purpose of the model.
study was to estimate the components of the behavior of Loh et al. [16] developed advanced statistical methods
the dam and determine the significance, extent and type for extracting trends from long term structural health
of irreversible displacements for 9 years. The hydrostatic monitoring data and try to set an early warning threshold
season time (HST) model was used to perform the sta- level based on the results of the analyses of Fei-Tsui dam in
tistical analysis. Using principal component analysis, Taiwan. The daily-based dam deformation data at one
investigations of dependencies between displacements sample per day and the temperature data were measured for
over the entire structure were done. Results showed that a period of 22 years. Data was analyzed using the singular
the thermal effects, response to reservoir water level and spectrum analysis with auto regressive model (SSA-AR)
the irreversible effects are the three main components of and the nonlinear principal component analysis (NPCA)
displacements. using auto-associative neural network method (AANN). It
De Sortis and Paoliani [13] compared statistical and was concluded that the AANN method could capture the
structural identification techniques in dam monitoring. Air periodic variations (temperature, season) as well as the
mean daily temperature, reservoir water level and crest trends (creep). This made the AANN model suitable for a
deformations spanning over 40 years were considered for long term trend prediction as compared to the SSA-AR
structural health assessment of Ancipa dam. Structural model which can only be applied for short time prediction.
identification was performed to obtain the relationship Popescu [17] carried out dam monitoring and surveil-
between external loads and crest displacements using lance using blind source separation on Vidraru dam in
numerical models. Results showed that the thermal con- Romania for a period of 1,200 days. This study was carried
tribution estimated with the structural identification pro- out with an objective of finding out contributions of the
cedure followed the air temperature cycles more closely external loads namely: air temperature and hydrostatic
than that estimated with statistical analysis, in which a pressure on structural deformation and separate these
sinusoidal behavior has been estimated. The study con- external loads of the dam and the time effects without prior
cluded that the structural identification procedure provided knowledge of the propagation environment. Second order
a higher degree of accuracy in predicting the future blind source separation (BSS) algorithm was used to
behavior of the structure and could be used in investigating determine the independent sources and identify the irre-
structural defects and ageing of dams. versible component in the structural response of the dam.
Perner and Obernhuber [14] analyzed the deformations Results indicated that the blind source separation method
of Zillergrund larch dam in Austria based on a hybrid was efficient in dam monitoring.
model which combines the structural analyses with statis- Henriques et al. [18] presented results of the collected
tical models. The hybrid model was developed based on data, namely inclination, temperature and reservoir level.
deformations measured for 19 years. Structural models The inclination measurements were taken in connection
were employed to identify the physical relations between with a test on a fully automated geodetic system for con-
deformations and water load and between the deformation tinuous monitoring of displacements. Using regression
and thermal load. Authors concluded that the hybrid model models, a good correlation was observed between the
had an advantage of having good predictions with only a variations in the inclination with temperature and with
small number of parameters for statistical analysis and that changes in the reservoir level of Cabril dam. Inclination

123
Author's personal copy
238 J Civil Struct Health Monit (2014) 4:235244

measurements were conducted for 2 days during which and STFT analysis. This shows that there is advancement
temperature was also measured on the downstream of the in the area of static monitoring of dams all over the world.
face of the dam. The study showed that temperature and Table 1 summaries the methods of analysis, their principle,
water level had an influence on the structural behavior of advantages and limitations plus the name and type of dam
the dam. Authors noted that there were some variations in in static monitoring.
the components of the inclination, which cannot be
explained by the regression models used.
Demirkaya and Balcilar [19] compared two statistical 3 Dynamic monitoring of dams
methods namely multiple linear regression (MLR) and
multiple layer perceptron (MLP) models to construct a As part of ICOLD recommendations, the performance of
daily displacement forecasting system of Schlegeis arch dams under vibrations has to be evaluated [21]. These
dam. Benchmark data were water level, air temperature and vibrations are may be due to wind, water waves and ground
concrete temperatures at six pointsone value per day for motions. Structural properties such as stiffness, damping of
8 years. The observed displacements of the dam were dams are affected by these vibrations. Although these
analyzed using time histories of water level and concrete properties can be modelled using finite element methods,
temperatures as input parameters. According to the ana- the real behaviors of these structures need to be verified
lysis, MLR exhibited a better performance under the cri- using field tests. Field tests facilitate the estimation of
teria of R2 as compared to MLPs R2. Results indicated that dynamic properties also commonly known as modal
linear regression provided the most appropriate solution of parameters namely: resonant frequencies, damping ratio
linear problems and acceptable in static monitoring of and mode shapes. There exist basically two field tests
dams. available for testing the behavior of dams under vibrations.
Mata [20] identified the effect of the daily variation of These include (1) forced vibration testing (FVT) and (2)
air temperature on the structural response of Alto Lindoso ambient vibration testing (AVT).
dam. The Short Time Fourier Transform (STFT) method In the forced vibration tests, dams are excited by
was used to identify the influence of the daily variation of mechanical means using shakers and the vibration
air temperature on the horizontal displacement of concrete responses are measured by accelerometers. With the force
dams and track its variation as a function of time which can signal acting as a reference to all response signals, dynamic
be used to validate new observations. Data used in this properties of the dam are extracted using frequency
study corresponded to a period between October 2008 and response functions (FRFs). FVT were the first field tests to
September 2011 with samples collected every hour. The be conducted on dams with the aim of understanding the
amplitude of the horizontal displacement was modelled as dynamic behavior of dams, validation and calibration of
a function of the amplitude of air temperature by a linear numerical models. For example, Severn [22] conducted
regression model after a strong correlation was verified. forced vibration tests on Wimbleball buttress dam with the
Results showed that a better correlation between the aim of determining the effect of rising water level to the
magnitudes was obtained when the phase of the horizontal structural behavior of dams. Artificial excitation was
displacement is more regular. caused by rotating eccentric mass (REM) exciters placed at
In summary, static monitoring of dams has been used for the crest of the dam. Resonant frequencies in the range
decades with the aim of investigating the contribution of 8.99.6 Hz were estimated from resonant peaks. Results
external loads to structure deformation and identify irre- indicated that the rise in water level caused a fall in reso-
versible components in the structural response. Parameters nant frequencies and the buttresses did not act indepen-
such as ambient temperature, displacements, reservoir dently from each other for small amplitudes of motions
water level to mention but a few. A lot of data is obtained generated by the exciters. Clough et al. [23] used field
from the measurements which require interpretation to get measurements in the updating of finite element model of
meaningful information. Statistical and deterministic Xiang Hong Dian dam. Forced vibration tests were carried
models have been developed to analyze the large data out using eccentric mass shakers. Accelerometers were
obtained from the monitoring programs. Statistical models placed on both the dam crest and the foundation on the
have been widely used as seen in the literature because they downstream face of the dam. Then 12 fundamental modes
are simple in formulation and able to provide a correlation in the range 412 Hz were obtained from frequency
between correlation between governing and dependent response curves. The authors concluded that field mea-
parameters. Examples of these models include HST, mul- surements correlated well with analytical predictions.
tivariate statistical analysis (MLR, PCA, ANN, and SSA- Cantieni et al. [24] created and updated a finite element
AR). Other techniques that have been used in data inter- model using field results on Norsjo dam in Sweden. The
pretation of static monitoring data recently include BSS servo hydraulic shaker was used to mechanically excite the

123
Author's personal copy
J Civil Struct Health Monit (2014) 4:235244 239

Table 1 Methods of analysis, features and name of the dam as applied in static monitoring
Method of analysis Principle Advantages Limitations Name (type) of dam

Deterministic model Compares the actual behavior of the Used for long term Theoretical Ferden arch dam Roggiasca
structure to the theoretical monitoring of the model may not arch dam
behavior structure be correct
Hydrostatic-season- Takes into account the hydrostatic Easy to implement useful Dependence on Idduki arch dam Daniel
time model level as a fourth degree in small dams where the the cyclic Johnson arch dam
polynomial, the seasonal effects as effect of the evolution of the
a sum of four trigonometric temperature in the external
functions and irreversible effects displacements is thermal loads
as a combination of exponential
and polynomial functions of time
Discrete fourier Decomposes time series signal into Gives information about Data has to be Landon gravity dam
transform frequency components each having periodic components uniformly
amplitude and phase distributed
Lomb normalised Estimates the frequency spectrum Can handle long periodic Time consuming Landon gravity dam
periodogram based on least squares fit of noise in long gapped
sinusoids to data samples records
Multiple linear Models the linear relationship Simple to use Speed of Does not take Ancipa gravity dam Schlegeis
regression between dependent variable and execution into account arch dam Cabril arch dam
various independent variables non-linearity Idduki arch dam
Zillergreundl arch dam Alto
Rabago arch dam
Short time fourier Breaks the signal into small time Can handle non stationary Requires Alto Lindoso arch dam
transform segments and performs Fourier signals experienced
analyses of each time segment to user
ascertain frequencies that exist in it
Multi-layer Based on forward artificial neural Detects trends that are Requires Schlegeis arch dam
perceptron network model that maps sets of complex to see experienced Alto Rabago arch dam
input data onto a set of appropriate user
output
Blind source Separates a set of source signals Low computation Source of Vidraru arch dam
separation from a set of mixed signals without complexity ambiguity
aid of information about the source
signals
Principal component Uses the dependencies between Identifies patterns Sensitive to noise Daniel Johnson arch dam
analysis variables to represent it in a more Data compression Idduki arch dam
tractable, lower-dimensional form
without losing much information
Singular spectrum Based on multivariate statistics of Eliminates the effect of Time consuming Fei Tsui arch dam
analysis with auto decomposing a signal into a sum of noise
regressive model several principal component time
series
Nonlinear principal Multivariate data analysis method Handles nonlinear Computationally Fei-Tsui arch dam
component analysis that uses dimensionality reduction, relationships between expensive
using auto- visualization, and exploratory data variables Compresses
associative neural analysis to identify and remove data
network method correlations among problem
variables
Structural Based on experimental Higher degree of Does not Landon gravity dam
identification measurements in terms of an accuracy in predicting consider non-
technique inputoutput data stream of the future linearity
unknown system are used to model
the system mathematically

dam. Dam responses in term of accelerations were mea- For more literature on forced vibration testing of dams see
sured in three directions at the dam crest. Using the least Refs. [25, 26].
square complex exponential (LSCE) algorithm, resonant Dams tend to be located in remote areas where acces-
frequencies in the range 3.5512.9 Hz were determined. sibility might be a problem and the transportation of the

123
Author's personal copy
240 J Civil Struct Health Monit (2014) 4:235244

mechanical exciters to site become very complicated and Kemp [32] conducted ambient vibration tests on a Ru-
expensive. In such cases, AVT becomes the only practical skin dam as part of safety evaluation studies of concrete
means of carrying out dynamic tests on dams. In ambient gravity dams in British Columbia, Canada. Dynamic
vibration testing also known as operational modal analysis, properties of Ruskin dam were used to calibrate a numer-
only responses from the structure are measured and it uses ical model of the dam. Natural frequencies in the range
environmental forces (wind, ground motion) to excite the 6.514 Hz for the high reservoir and 8.514.5 Hz for low
structure. Instead of using the force signal as a reference, reservoirs were obtained from the peaks of average nor-
the response signal measured at one or more points (often malized power spectral densities (ANPSDs).
called reference points) is used while estimating the Danielle and Taylor [9] presented results of ambient
dynamic properties. AVT has advantages over FVT, i.e. it vibration tests for purposes of validating a finite element
is cheaper and the structure is tested in its operating con- model for the dam-reservoir foundation system of Claew-
ditions. Further developments in data acquisition and effi- ern dam. Force balanced accelerometers mounted on steel
cient methods in either frequency and time domain blocks on the dam crest were used to acquire the signals.
methods available to process signals from ambient vibra- Natural frequencies were picked from the peaks of average
tion tests [27, 28] has made AVT a preferable method in normalized power spectral densities. Six lateral modes
SHM of dams. were identified in the frequency range 6.111.8 Hz. The
The first reported attempt to conduct ambient vibration computed mode shapes and natural frequencies from the
tests on dam was on Contra dam in Switzerland [29]. The finite element analysis compared well with the experi-
objective of the study was to investigate the feasibility of mental results. The study demonstrated that ambient
ambient vibration testing of dams. Force balanced acceler- vibration testing can offer a viable alternative to forced
ometers placed on the crest of the dam were used to measure vibration testing when only the modal properties of a dam
vibrations. Natural frequencies were determined by picking are required.
peaks on the auto-spectrum while the mode shapes where Darbre et al. [33] investigated the effect of various
obtained from the output only transfer functions between the reservoir levels on natural frequencies of the arch dam of
travelling and reference accelerometers. Eight upstream Mauvoisin in Switzerland. Ambient vibration tests on
downstream modes were extracted in the frequency range Mauvoisin dam marked the beginning of long term ambient
1.84.2 Hz. Test results showed that the natural frequencies vibration monitoring of dams. Seven ambient vibration
of the measured models increased with the decline in the tests at different water levels were carried out between
reservoir level and the excitation process was directly related 1995 and 1996. An automated system was set up on the
to hydroelectricity generating activity. dam and ambient vibrations were recorded twice daily for
Brownjohn [30] investigated the safety and stability of 6 months. Natural frequencies were obtained by peak
Hermitage dam, a concrete gravity dam located on the Wag picking on normalized power spectral densities of indi-
River, Kingston, Jamaica. Field results were used to vali- vidual acceleration. Results showed that the natural fre-
date a finite element model of the dam. Natural frequencies quencies initially increased with rising water level and then
in the range of 730 Hz picked from peaks of the auto- decreased with further rise. This was contributed to the two
spectra were compared with the analytical natural fre- competing features of increasing mass of water (reduction
quencies obtained from the finite element model of the of natural frequencies) and of dam stiffening due to closing
dam. The report concluded that although there were some of vertical construction joints.
discrepancies between the predictions of the mathematical Mivehchi and Ahmadi [34] carried out ambient vibra-
model and the experimental results, the agreement was tion tests on Shahid-Rajaee and Saveh dams, two concrete
good enough to justify the usage of the mathematical arch dams in Iran. The purpose of the tests was to verify the
model. results obtained from mathematical model used regularly in
Loh and Wu [31] obtained the dynamic characteristics the Iranian dam design practice by comparing with the
of the dam using seismic response data and ambient behavior of the actual as-built structures. The ambient tests
vibration data of Fei-Tsui dam in Taiwan. The modal were conducted during winter of 1999 to autumn of 2000.
parameters were extracted for the upstreamdown direc- Partial and rapid opening and closing of the bottom outlet
tion. Random decrement method and auto regressive model gates of the dam body provided artificial excitation during
with least-squares method were used to estimate modal the ambient vibration tests. These were helpful in exciting
parameters of the dam. Two modes were identified with the any weak modes which could not be realized with ambient
first and second having average frequencies of 2.49 and tests only. Modal parameters from both dams were
3.33 Hz. Results showed that the identified natural fre- extracted from peaks of auto spectrum. Natural frequency
quencies from ambient vibration tests verified that were ranges of Shahid-Rajaee and Saveh dams were reported as
consistent with those from earthquake excitation. 1.463.58 and 3.917.91 Hz, respectively. Authors

123
Author's personal copy
J Civil Struct Health Monit (2014) 4:235244 241

concluded that ambient vibration tests can be used in the monitoring. However, authors concluded that ambient
seismic safety evaluation of other Iranian dams and also be vibration tests add value to safety surveillance of dams.
used in the design enhancement of new dams. Ellis et al. [43] reported ambient vibration tests which
In addition, a continuous dynamic monitoring system were performed on Gem Lake dam in California, USA. The
was pioneered at Cabril arch dam in Portugal. This was objective of the tests was to measure the dynamic response
after ambient vibration tests were carried out on this dam characteristics of the entire dam. Spectral analysis tech-
between 2000 and 2003 [35]. The results from these two niques that included the fast Fourier transform and the
tests were used for different purposes (1) demonstrate that maximum entropy method, coupled with waterfall plot
it is possible to characterize the dynamic response of arch analyses, were effective in the identification of individual
dams with good precision owing to the development of arch and full dam model behavior. Natural frequencies of
sensors and data acquisition systems [36], (2) investigate the arches ranged from 13.18 to 27.71 Hz. Numerical
the influence of contraction joints on dynamic behavior of modelling of the dam was carried out to gain an insight of
Cabril dam [37] and (3) study the influence of an intake how the dam behaves and was used to develop a suitable
tower dynamic behavior on modal identification of the dam field test procedure. It was concluded that a practical field
[38]. Frequency domain decomposition (FDD) and sto- test approach based on ambient vibration testing can be
chastic subspace identification (SSI) techniques were used adopted for large multiple arch dams.
to extract modal parameters of the dam and also to cross- Sevim et al. [44] presented finite element model cali-
validate the results. Numerical results from 3D finite ele- bration of Berke arch dam in Turkey using operational
ment models were compared with the experimental results. modal analysis. Ambient vibration tests were conducted on
Results showed that there was a need to develop a con- the dam to obtain the vibration characteristics. Wind and
tinuous monitoring system for the dam so that the finite water pressure were the source of excitation during ambient
element models would be calibrated. The performance of vibration testing of Berke dam. Using enhanced frequency
the different modal estimation methods on Cabril dam was domain decomposition modal parameters of the dam were
not evaluated. obtained. Eight natural frequencies ranging between 2.74
Okuma et al. [39] collected fundamental data for and 9.66 Hz were determined. The analytical finite element
developing structural damage detection based on long term model of Berke arch dam was calibrated to match with the
ambient vibration testing of Hitotsuse arch dam in Japan. measured dynamic characteristics of the dam.
Long term vibration testing was done to evaluate the In summary, dynamic monitoring of dams using ambient
changes of the natural frequencies of the whole dam due to vibration tests has been going for the last two decades. The
macroscopic damage caused by aging effects. The mea- major aim of all the tests was to determine the dynamic
suring system was configured to record continuously at a characteristics of dams and calibrating mathematical
sampling rate of 200 Hz and store data measured every models. Other aims included: studying the effect of water
after 30 min. Natural frequencies were identified from the reservoir levels to the resonant frequency, structural health
cross spectrum of autoregressive moving average models. monitoring and for seismic evaluation studies. In the
Results showed that the identified natural frequencies were extraction of dynamic properties from the ambient data
in good agreement with the earthquake observation from the dams, frequency domain procedure (peak picking)
records, and the identified natural frequencies of three was widely used. This is because of its simplicity in the
modes strongly correlated with the water level of the dam. extraction of the dynamic properties. Other methods used
Moyo and Oosthuizen [40] performed ambient vibration include curve fitting of the transfer functions and random
survey trials on two arch dams, namely, Roode Elsberg and decrement method. However, with the development in
Kouga located in South Africa, respectively. The major signal analysis techniques, new methods such as frequency
aim of the tests was to obtain dynamic characteristics of domain decomposition and stochastic subspace identifica-
these dams to be used as baseline measurements for long tion have also been used in the analysis of raw data from
term dam safety monitoring. Four sets of ambient vibration ambient vibration tests of dams. Table 2 below presents a
measurements were taken on Roode Elsberg dam between summary of the methods of analysis, their principle,
a period of December 2008 and April 2010 while only one advantages and limitations followed by the name and type
measurement test in September 2009 was carried out on of dam for which these were applied.
Kouga dam. Several operational modal analysis (OMA)
methods have been compared using test data from these
two dams [41, 42]. Natural frequencies of Roode Elsberg 4 General summary
dam and Kouga dam were in the range 37.70 Hz and
3.728.30 Hz, respectively. Results showed that there was Monitoring the static and dynamic behavior of large dams
no best OMA algorithm that was suitable for dam is a topic of great relevance, due to the impact that these

123
Author's personal copy
242 J Civil Struct Health Monit (2014) 4:235244

Table 2 Methods of analysis, principles and name of the dam as applied in dynamic monitoring
Method of Principle Advantages Limitations Name (type) of dam
analysis

Curve fitting of Based on curve fitting of frequency transfer Simple to Picking the peaks is always a Contra arch dam Fei-Tsui
frequency functions obtained by Fourier analysis. use subjective task; damping arch dam Gem Lake arch
transfer The computed transfer functions are used estimates are unreliable dam
functions to estimate dynamic properties of a
structure
Peak picking The auto-spectra are normalized and Simple to Picking the peaks is always a Ruskin gravity dam
averaged in order to obtain an averaged use subjective task; Damping Claewern gravity dam
normalized power spectral density estimates are unreliable. Mauvoisin arch dam
function that, in principle, shows all the Cannot handle noise Hitotsuse arch dam Saveh
resonance peaks corresponding to the efficiently arch dam
vibration modes of a system
Frequency Power spectral densities of a system are Detects Cannot determine damping Cabril arch dam Berke arch
domain evaluated by performing singular value closely ratios dam
decomposition decomposition of the response spectral spaced
matrix. The spectral densities are modes
decomposed in the contributions of the
different modes of a system that, at each
frequency, contribute to its response
Stochastic Based on the stochastic subspace model Detects Stabilization diagram has many Cabril arch dam Roode
subspace theory. It identifies the state matrix and closely parameters Elsberg arch dam
identification output matrix which contains dynamic spaced
properties of the structure using QR- modes No
factorization and singular value leakage
decomposition
Random Time domain procedure, where the No leakage Requires experience Fei-Tsui arch dam
decrement structural responses are transformed into and less
method random decrement functions, which are random
proportional to the correlation functions of
the system operational responses

structures have on the landscape where they have been interpretation in static monitoring, long time data records
built. Health monitoring of dams plays an important role in are needed. This can be seen that in the presented examples
securing their structural integrity and maintaining their most data are beyond 10 years although there is one for
longevity. This is divided into two namely: static and \4 years given in the review. The review shows that static
dynamic monitoring. monitoring of dams has an excellent track record and is
The paper has presented a literature review on the case very useful in the assessment of dams health.
studies on health monitoring of dams using static and Dynamic monitoring of dams has been going on since
dynamic properties. Static monitoring of dams involves 1980s with forced vibration testing being the first to be
measurement of static factors such as ambient tempera- conducted on dams. Short term measurements of dynamic
tures, reservoir level, opening and closing of joints, crack properties were done to mostly determine the relationship
opening, displacements and strains which are measured between reservoir water level and resonant frequencies,
accurately by instruments. Mostly statistical models (HST then validation and calibration of analytical models. From
model and multivariate statistical methods) have been used the year 1990, ambient vibration testing became a prefer-
in the interpretation of field data although there were other able methodology due to the advancement in technology
methods implemented as structural identification tech- for data acquisition systems and research in signal pro-
niques, short time Fourier transform and blind source cessing techniques. In the earlier years, only once off
separation. The main reason why statistical methods have measurements from AVT of dams were used for several
been widely used is that they are simple in formulation and purposes such as; (1) investigation of the feasibility and
providing any correlation between governing and depen- suitability of AVT of dams (2) investigating the safety and
dent parameters. In all the presented examples a lot of stability of dams (3) determine dynamic parameters for
focus was put on the effect of reservoir water level, tem- calibration and validation of finite element models of dams.
perature on deformations and most of them have been a However, from the year 2000, repeated measurements have
success. It can also be seen that for meaningful data been carried out on different dams for; (1) investigating the

123
Author's personal copy
J Civil Struct Health Monit (2014) 4:235244 243

effect of reservoir levels on natural frequencies (2) devel- those continuously extracted from long term dynamic
opment of a continuous dynamic monitoring system and monitoring systems so as any structural anomalies can be
(3) development of damage detection system based on long detected.
term AVT. This shows that AVT can now be used for long There is also a need to integrate the static response data
term safety monitoring of dams. For most of the presented with data from dynamic monitoring. Comparison of the
case studies, frequency domain based (peak picking and evolution of dynamic properties especially natural fre-
frequency domain decomposition) have been used to quencies and say deformations with time. This will be used
extract the dynamic properties from vibration measure- to further provide an indication of the full performance of
ments. This is because these methods are simple and easy the dams.
to use as compared to the time domain based methods. Development of new analysis tools which can provide
threshold values of structural responses (deformation and
natural frequencies). This will help dam safety engineers to
5 Critical issues for future research in health find solutions in case these threshold values are reached in
monitoring of concrete dams the evaluation of the dam.
In conclusion, despite rigorous and considerable
This section contains a summary of the critical issues as research in monitoring the structural integrity of infra-
perceived by the authors in the field of health monitoring structure notably in bridges and buildings, health moni-
of concrete dams. The importance of this section is to toring of dams is still lagging behind. While dam
focus on the issues that need to be addressed by future monitoring practice has been in existence for longer than
research to make health monitoring of concrete dams other civil engineering disciplines such as bridges, this
more feasible. subject in dams has not kept pace with recent advances in
The literature has little instances of studies where dif- sensing technologies and data analysis methodologies. If
ferent data analysis methods in static monitoring are specific methodologies can be implemented to evaluate the
compared directly by application to a common data set. structural health of concrete dams and extend the life of
There are a few case studies [15, 19] where different these structures, the investment in this research will be
methods of analysis in static monitoring have been pre- worthwhile.
sented, however the relative merits of these methods and
their success in static monitoring of dams have not been Acknowledgments The authors would like to acknowledge the
financial support of Water Research Commission (WRC) South
directly compared in a sufficiently objective way. Africa of the on-going project. The authors also thank the Department
Secondly, results from dynamic tests can be influenced of Water Affairs (DWA) South Africa for accepting the access of the
by the following factors: fluidstructure interaction, foun- dams for monitoring. Support provided by the Concrete Materials and
dation deformability and non-linear behavior of cracks Structural Research Unit (CoMSIRU) is highly recognized.
which can change the dynamic properties (natural fre-
quencies) of the structure. Therefore, research should not References
be limited to water level effects on natural frequencies but
also to the above mentioned factors. 1. Graham WJ (1999) A procedure for estimating loss of life caused
In addition, an issue that has received almost no atten- by dam failureDSO-99-06. Bureau of Reclamation, Denver
tion in dynamic testing of dams is damage detection. That 1999
2. Rich TP (2006) Lessons in social responsibility from the austin
is to say the ability to distinguish between changes in the dam failure. Int J Engng Ed 22(6):12871296
dynamic properties resulting from damage and those from 3. Brownjohn JW (2007) Structural health monitoring of civil
environmental and operational loads. Several algorithms infrastructure. Phil Trans R Soc A 365:589622. doi:10.1098/
have been applied to other structures such as bridges and rasta.2006.1925
4. ANCOLD (1994) Guidelines on risk assessment. Australian
buildings but these cannot be applied to dams. This is National Committee on Large dams, Sydney
because real-life vibration monitoring of dams are different 5. Reynolds S (2013) Evaluating the decision criteria for the pri-
from bridges and buildings. Therefore there is a need for oritization of South African dams for rehabilitation in terms of
research on the damage detection algorithms in dam risk to human lives. Dissertation Stellenbosch University
6. ICOLD (1987). Dam safety: guidelines, Bulletin 59. Technical
engineering. Report, International Committee on Large Dams
Finite element analysis in dynamic monitoring using 7. Bianch M, Bremen R (2000) Health monitoring of arch dams
ambient vibration tests is another area that has not been recent developments. In: Proceedings of the International
fully explored. The case studies presented in this paper Workshop on the Present and Future in Heath Monitoring.
Bauhaus
only compared analytical results to experimental results [9, 8. Behrouz AN (2002) Multivariate statistical analysis of moni-
30, 32, 38, 44]. However, there is a need to integrate the toring data for concrete dams. Dissertation. McGill University,
information provided by finite element modelling with Montreal

123
Author's personal copy
244 J Civil Struct Health Monit (2014) 4:235244

9. Danielle WE, Taylor CA (1999) Effective ambient vibration Proceedings of the 25th International Modal Analysis Confer-
testing for validating numerical models of concrete dams. Earthq ence, Orlando
Eng Struct Dynam 28:13271344 29. Brownjohn JMW, Severn RT, Taylor CA (1986) Ambient
10. Zhang LM, Brinker R, Andersen P (2005) An overview of vibration survey of Contra dam. Research Report University of
operational of modal analysis: major developments and issues. In: Bristol Department of Civil Engineering, UK
Proceeding of 1st International Operational Modal Analysis 30. Brownjohn JMW (1990) Dynamic investigation of Hermitage
Conference, Copenhagen Dam, Jamaica. Report UBCE-EE-90-13 University of Bristol
11. Pytharouli S, Iand Stathis CS (2005) Deformation and reservoir Department of Civil Engineering, UK
level fluctuations: evidence for a causative relationship from 31. Loh CH, Tsu-Shiu W (1996) Identification of Fei-Tsui arch dam
spectral analysis of a geodetic monitoring record. Eng Struct from both ambient and seismic response data. Soil Dyn Earthq
27:361370 Eng 15:465483
12. Chouinard L, Lariviere R, Cote P, Zhao W (2006) Analysis of 32. Kemp BG (1996) Ambient vibration assessment of Ruskin dam:
irreversible displacements in multiple arch concrete dam using dynamic properties. University of British Columbia (UBC),
principal component analysis. In: The Proceedings of the Inter- Vancouver
national Conference on Computing and Decision making in Civil 33. Darbre GR, De Smet CAM, Kraemer C (2000) Natural fre-
and Building Engineering, Montreal quencies measured from ambient vibration response of the arch
13. De Sortis A, Paoliani P (2007) Statistical analysis and structural dam of Mauvoisin. Earthq Eng Struct Dyn 29:558577
identification in concrete dam monitoring. Eng Struct 29:110120 34. Mivehchi MR, Ahmadi MT (2003) Effective techniques for arch
14. Perner F, Obernhuber P (2010) Analysis of arch dam deforma- dam ambient vibration test. Application on two Iranian dams.
tion. Front Archit Civ Eng China 4(1):102108. doi:10.1007/ Report from the International Institute of Earthquake and Seis-
s1170-010-0012-7 mology (IIEES), Tehran
15. Mata J (2011) Interpretation of concrete dam behavior with 35. Oliviera S, Rodrigues J, Mendes P, Costa AC (2004) Damage
artificial neural network and multiple linear regression models. characterization in concrete dams using output-only modal ana-
Eng Struct 33:903910 lysis. In: The Proceedings of 22nd International Modal Analysis
16. Loh CH, Chen CH, Hsu TY (2011) Application of advanced Conference Dearborn, Michigan
statistical methods for extracting long-term trends in static 36. Mendes P, Oliveira S, Guerreiro S, Baptista MA, Campos Costa
monitoring data from an arch dam. J Struct Health Monit. doi:10. A (2004) Dynamic behavior of concrete dams monitoring and
1177/1475921710395807 modelling. In: The Proceedings of the 13th World Conference on
17. Popescu TD (2011) A new approach for dam monitoring and Earthquake Engineering, Vancouver
surveillance using blind source separation. Int J Innov Comput 37. Mendes P, Oliveira Costa C, Almeida Garrett J, Oliveira S (2007)
Inf Control 7(6):38113824 Development of monitoring system to Cabril dam with opera-
18. Henriques MJ, Lima JN, Oliviera S (2012) Measuring inclina- tional modal analysis. In: The Proceedings of the 2nd Experi-
tions in Cabril dam with an optoelectronic sensor. In: Proceedings mental Vibration Analysis for Civil Engineering Structures
of the FIG working week 2012, Rome (EVACES), Porto
19. Demirkaya S, Balcilar B (2012) The contribution of soft com- 38. Mendes P, Oliveira S (2009) Influence of the intake tower
puting techniques for the interpretation of dam deformation. In: behavior on modal identification of Cabril dam. In: The pro-
Proceedings of the FIG working week 2012, Rome ceedings of the 3rd International Operational Modal Analysis
20. Mata J, Tavares de Castro A, Sa da Costa J (2013) Timefre- Conference Portonovo
quency analysis for concrete dam control: correlation between the 39. Okuma N, Etou Y, Kanzawa K, Hirata K (2008) Evaluation of
daily variation of structural response and air temperature. J Eng dynamic properties of an aged large arch dam. Civil Engineering
Struct 48:658665 Group. Research Laboratory report, Kyushu Electric Power Co.,
21. Wieland M, Kirchen GF (2012) Long-term dam safety monitor- Fukuoka
ing of Punt dam Gall arch dam in Switzerland. J Struct Civil Eng 40. Moyo P, Oosthuizen C (2010) Ambient vibration survey trials of
6:7683 two arch dams in South Africa. In: The Proceeding of the 8th
22. Severn RT, Jeary AP, Ellis BR (1980) Forced vibration tests and ICOLD European Club Symposium, Innsbruck
theoretical studies on dams. In: Proceedings of Institution of Civil 41. Bukenya P, Moyo P, Oosthuizen C (2012a) Modal parameter
Engineers, Part 2, vol 69 ICE London estimation from ambient vibration measurements of a dam using
23. Clough RW, Chang KW, Stephen RW (1986) Vibration behavior stochastic subspace identification methods. In: The Proceedings
of Xiang Hong Dian dam. UC, Berkeley of the 3rd International Conference on Concrete Repair, Reha-
24. Cantieni R, Wiberg U, Pietrzko S (2004) Modal investigation of a bilitation and Retrofitting Cape Town, South Africa
dam. In: Proceedings of the 16th International Modal Analysis 42. Bukenya P, Moyo P, Beushausen H, Oosthuizen C (2012b)
Conference, Santa Barbara, 1998: 11511157 Comparative study of operational modal analysis techniques
25. Deinum PJ, Dungar R, Ellis BR, Jeary AP, Reed GAL (1982) using ambient vibration measurements of a concrete dam. In: The
Vibration tests on Emmoson arch dam, Switzerland. Earthq Eng Proceedings of the 25th International conference on Noise and
Struct Dynam 10(3):996 Vibration Engineering, Leuven
26. Duron ZH, Hall JF (1988) Experimental and finite element 43. Ellis E, Duron Z, Von Gersdorff N, Knarr, N (2010) Dynamic
studies of the forced vibration response of Morrow Point dam. characterization of a large multiple arch dam. In: The Proceed-
Earthq Eng Struct Dynam 16(7):10211039 ings of the 30th Annual USSD conference, Sacramento
27. Van Overschee P, De Moor B (1996) Subspace identification for 44. Sevim B, Bayraktar A, Altunisk AC (2010) Finite element model
linear systems: theoryimplementation and applications. Kluwer calibration of Berke arch dam using operational modal testing.
Academic Publishers, Dordrecht J Vibra Contr. doi:10.1177/1077546310377912
28. Brincker R, Andersen P, Jacobsen NJ (2007) Automated fre-
quency domain decomposition for operational modal analysis. In:

123

View publication stats

Potrebbero piacerti anche