Estimation of structural parameters using static loading tests

E. Covián & M. Casero

Escuela Politécnica de Mieres, Universidad de Oviedo, Asturias, Spain
A. González
School of Civil, Structural and Environmental Engineering, University College Dublin, Dublin Co., Ireland

ABSTRACT: The increasing stock of aging infrastructure demands new convenient and efficient meth-
ods for early damage detection. In this paper, focus is placed upon methods that use static measure-
ments via modern surveying techniques. Theoretical simulations play an important role in the devel-
opment of damage detection methods. These simulations use noise-free data or corrupt the ideal
response with noise to resemble a field scenario. However, the nature and true uncertainty of the
measured structural response is commonly ignored. Here, the uncertainty associated to modern sur-
veying techniques employed for measuring deflections and rotations is quantified and used to propose
site-specific noise models. The latter are used to assess the accuracy of a Cross-Entropy static damage
detection algorithm.

these measurements affect the accuracy of these

1 INTRODUCTION techniques.

An increasing number of large structures are ap- For a given load, the static response is related to
proaching the end of their design life, while oth- the boundary conditions and distribution of
ers may have been built based on wrong assump- structural stiffness throughout the structure. As
tions (i.e., traffic flows underestimated in damage can be characterized as a local reduction
bridges), which could represent an aging acceler- in structural stiffness, analysis of the structure’s
ation factor. Health monitoring of these struc- static response has been used extensively to lo-
tures is required to ensure their safety. There is cate and quantify damage (Sanayei & Scampoli,
now technology to live-monitor characteristics of 1991; Hjelmstad & Shin, 1997; Bakhtiari-Nejad et
a structure such as acceleration, deflection, rota- al., 2005; Caddemi & Morassi, 2007, Walsh &
tion, strain and temperature and to provide an Gonzalez 2009 and Yang & Sun, 2010). Most of
accurate image of the structure’s health. Howev- these methods use a finite element (FE) mathe-
er, this monitoring is often overlooked for many matical model as reference, and deflections and
structures due to cost and design/installation rotations as inputs. However, these inputs need
complications. Therefore, there is a need for a to be measured and “no measurement is ever ex-
convenient and efficient method of monitoring, so act” (Ghilani & Wolf, 2006). In order to address
that damage can be detected sooner and propaga- the later, preliminary theoretical testing of new
tion minimized. This paper focuses on damage algorithms may pollute the ideal structural re-
detection techniques based on static measure- sponses with random errors in an attempt to
ments, and how the number of measured points simulate a more realistic scenario. However,
(measurement density) and the uncertainty of these simulated errors are not in agreement with
their own nature, and their uncertainty for a par- ter that would help to make that choice
ticular confidence level is commonly ignored. (Bakhtiari-Nejad et al. 2005; Yang & Sun, 2010).

Although previous researchers (Marchamalo et A review on dynamic based methods can be

al. 2011; de Luis, 2009; Gordon & Lichti, 2007; found in Yan et al. (2007), while an example of a
Olalla, 2007; Psimoulis et al. 2006; Li et al. 2006; method combining both static and dynamic data
Boehler & Marbs, 2004; Mills & Barber, 2004 and is provided by Wang et al. (2001). This paper fo-
Cheng et al. 2002) have investigated the accuracy cuses on static based methods and how meas-
of measurements for many techniques, this accu- urement uncertainty can affect their perfor-
racy has often been expressed as nominal values mance.
or referred to observable measurands (as op-
2.2 Damage Detection Methods based on static
posed to objective measurands), and as a result,
test data
the suggested values are not comparable. In this
paper, the uncertainty associated to modern sur- Sanayei & Scampoli (1991) propose a method to
veying instruments and techniques, such as pro- calculate structural parameters from a subset of
grammable and reflector-less total station and degrees of freedom, for which displacement
terrestrial laser-scanner is quantified. Then, noise measurements have been taken, by minimizing
models are proposed for simulating different field the error resulting from the comparison between
scenarios. Finally, the impact of measurement two stiffness matrix (analytical and measured).
uncertainty and measurement density on a static They analyse the acceptable level of error in the
damage detection algorithm is assessed using measurements to maintain the input/output er-
theoretical simulations corrupted with the noise ror relationship in the linear range. Once estab-
models. lished, they conduct a simulation performing
probabilistic parameter identification with dif-
ferent subsets of parameters and affecting meas-
2 DAMAGE DETECTION METHODS
urements by the error previously calculated.
2.1 General Overview Hjelmstad & Shin (1997) develop a static parame-
ter estimation algorithm that includes an adap-
Damage detection methods can be classified in
tive parameter grouping scheme in order to deal
two main groups: those that subject the structure
with sparse data. Hjelmstad and Shin use the out-
to static loading and those that use measure-
put error proposed by Banan et al. (1994), which
ments from dynamic test data. A third category
is defined as the difference between computed
could be considered for those methods that com-
and measured displacements. In addition, a
bine both static and dynamic test data.
measured data perturbation scheme is presented,
The main differences between both approaches where estimated parameters are considered as
are: i) equations of dynamic equilibrium depend random variables and two damage indices (calcu-
on mass, damping and stiffness matrix whereas lated using the mean and standard deviation of
equations of static equilibrium depend solely on their distributions) are introduced.
the stiffness matrix; ii) High order of modal data,
Bakhtiari-Nejad et al. (2005) introduce an algo-
as well as certain mode shapes, that are needed in
rithm based on the comparison of the load vec-
some dynamic methods, can be difficult to obtain;
tors of the damaged and undamaged structure.
iii) In static methods, damage at certain locations
An error vector is formulated as the difference
of the structure can be concealed due to load
between both load vectors and then a minimiza-
paths associated with the load applied. The latter
tion problem is stated subjected to two con-
reflects the importance of adequately choosing
straints: a non-linear equality that reflects the
the applied load as well as the need for a parame-
measured change in displacement and an ine-
quality ensuring negative change in stiffness.
Moreover, they include selection criteria for load pressed as a percentage of the computed or ana-
cases, baseline values and measurements loca- lytical value (examples in the literature range
tions. The measurements error is found to have from 0,002% to 10%). However, it would be in-
no effect on the algorithm performance. Never- teresting to choose the magnitude of the noise in
theless, no particular noise treatment is pro- relation to the precision of the measurement de-
posed. Caddemi & Morassi (2007) develop vice used in the field. This simulation will fulfil
closed-form expressions for the determination of the purpose of testing the algorithm in realistic
the location and magnitude of cracks in beams, conditions.
subjected to different boundary conditions
(pinned-pinned, pinned-clamped and clamped-
3 ACCURACCY OF SURVEYING TECHNIQUES
free). Cracks are modelled as rotationary springs
of constant related to the severity of the damage. 3.1 Principles in deflection and rotation
The accuracy of the estimated crack location is measurements
found to decrease significantly for low levels of
One of the goals of this paper is to determine if
damage.
the accuracy associated to a measurement tech-
Yang & Sun (2010) approach the problem of lo- nique is sufficient to guarantee the correct per-
cating and quantifying damage in a decoupled formance of the CE static damage detection algo-
fashion. The flexibility disassembly theory is used rithm mentioned in Section 2.2. Deflections and
as the location algorithm, obtaining a damage lo- rotations are the inputs used on this algorithm
cation vector that identifies the damaged ele- and reflector-less electronic tachymeter (TAC)
ments. Then, damage is quantified for those ele- and terrestrial laser-scanner (TLS), commonly re-
ments through a perturbed stiffness parameter. ferred to as high definition survey (HDS), the
Abdo (2011) conducts a parametric study of the measurement instruments selected. Deflection
relationship between damage and displacement and rotation are indirectly measured, since ob-
curvature. He states that damage represents a servable measurands are acquired firstly and
discontinuity in the curvature that can be located then a model is applied to calculate them.
and quantified using Grey system theory to com-
As in every measurement, many error sources
pare the curvature of damaged and undamaged
bring differences between the true values and
structures. Measurement noise is found to have a
measures of a particular measurand (BIMP,
negligible effect on the results.
2012). The nature of known error sources should
Walsh & González (2009) apply cross-entropy be analysed and procedures applied to reduce the
(CE) to damage detection. They develop an itera- error, in order to reach goodness of measures.
tive algorithm that creates a certain number of These procedures are focused, firstly, on elimi-
trial beams and selects those with a deflection nating gross and systematic error components
similar to the one measured. Then, based on the and, then, on minimizing random error compo-
statistics of the selected best trial beams, a new nent in measurements (Covián & Puente, 2013).
sample is generated and the process is repeated
Here, the stages of instrument calibration, at-
until convergence is achieved. Noise treatment
mospheric corrections and, in the case of ta-
and its influence on the algorithm performance
chymeter, Bessel compensation are supposed to
are yet to be explored. In a further section, the
be applied to reduce systematic errors. The most
performance of this CE algorithm is tested in the
unfavourable combination of random error
presence of field-like noisy measurements. Re-
sources is considered when quantifying, in pre-
garding the measurement errors, the usual ap-
analysis, the accuracy of a given instrumentation
proach, when a numerical simulation is per-
(see Sections 3.2 and 3.3). Pre-analysis error es-
formed, is to consider an “artificial” random noise
timations are propagated in order to calculate the
to simulate real measurements. This noise is ex-
final error of the objective measurands — tions (ui) in many points (i = 1, 2… n) distributed
defections and rotations—. along a beam (Fig. 1). In Figure 1, A and B are the
supports of the beam under investigation (solid
Finally the value of the measurand (x) should be
thick line represents the initial position and solid
expressed, in accordance with Guide to the ex-
dashed line represents the lj loading case), E is
pression of uncertainty in measurement (BIMP,
the location of the measurement instrument and
2008), showing not only the reference value of
ai is the height of the instrument. x E y E z E and
the measurand but also its uncertainty and asso-
ciated confidence level: x A y A z A are the axes of the reference systems for
the measurement instrument and for the beam
x = x ±U x = x ± k × s x (1) analysis, respectively. Therefore, it is necessary
to apply a Helmert transformation to go from
where x is the reference value, Ux is the uncer-
measured data to structural data, in which the
tainty, x is standard deviation of the measurand
ordinate y becomes the same for all the points.
and k is a coverage factor (k > 1) that depends on
The observational measurands are: geometrical
the selected probability distribution. In accord-
distance between the origin of the measurement
ance with the Bureau International des Poids et
instrument and the viewed point (dg) and hori-
Mesures recommendations (BIMP, 2008), 95% of
zontal () and zenithal () angles. These measur-
confidence level (k = 1.96) of the normal distribu-
ands are considered for the first position of every
tion is adopted here. 0
point ( Pi ) corresponding to the initial condition
Previous research has reported on the accuracy l
of the structure and for a second position ( Pi j )
of different measurement techniques. Some
(Marchamalo et al. 2011; Olalla, 2007; and oth- resulting from applying a loading case (
ers) mention precision data as the nominal value "j = 1,2...m ).
given by the manufacturers, in accordance or not
While TAC is more accurate than HDS, the latter
with ISO rules, and referred to observable but not
measures much more points (higher density) than
to objective measurand. Recent investigations on
TAC. The next two equations show the relation
differential GPS (Marchamalo et al. 2011; Cheng
between observable measurands and the deflec-
et al. 2002; Psimoulis et al. 2006; Li et al. 2006
tion (Eq. 2) and between the estimated errors of
and others) are carried out in large scale struc-
observable measurands and the error of the de-
tures (breakwaters, suspension bridges and sky-
flection (Eq. 3):
scrapers) characterized by being slender and
flexible structures with large displacements. ui = dg0 × cos l 0 - dgj × cos l j
l l
(2)
While some research on HDS try to determine
( )
2
precision through comparison with reference s u = 2 × s 2 × cos2 l 0 + s l2 × d 0 sin l 0
i dg g
(3)
values of point positions in laboratory (short dis-
l
tances) (Gordon & Lichti, 2007 and Boehler & where d g0 , dgj , l 0 and l l j are described above, and
Marbs, 2004), others try to do it in real scenarios s u , s dg , s l are the estimated errors of deflec-
i
(long distances) comparing results to other tech-
tions, geometrical distances and zenithal angles,
niques (de Luis, 2009 and Mills & Barber, 2004).
respectively. Equation 2 can be deduced from
Generally, HDS real scenarios show precisions
Figure 1, and Equation 3 comes from the applica-
unable to capture deflections smaller than 15 mm
tion of variance-covariance propagation law
accurately.
(Ghilani & Wolf, 2006). Equation 3 has been ob-
3.2 Surveying techniques to measure deflection tained considering no variation between the first
0 l
Both techniques —TAC and HDS— are based on ( Pi ) and the second position ( Pi j ) of every point,
the same observational measurands and rela-
tional models in order to determine the deflec-
since the effect of this assumption in deflection where a (usually expressed in mm) and b (in
error ( s ui ) is negligible. ppm), determined in accordance with ISO 17123-
3 (ISO, 2001), are the error component derived
Prior to the resolution of Equation 3, it is neces- from the phase-meter precision and the error de-
sary to establish the estimated errors of observa- rived from the instability of the wavelength of the
ble measurands: geometric distance error ( s dg ) EDM system respectively (Covián & Puente,
and zenithal angle error ( s l ), which can be calcu- 2013); comp is the precision of the vertical com-
pensator and ang is the angular error component
lated using Equations 4, 5:
in accordance with ISO 17123-4 (ISO, 2001). This
s d = a + b(ppm)× dg
g
(4) info is generally specified in the equipment
datasheets for k = 1 (68,26 % of confidence level),
(s )
2
sl = comp 3 + s ang
2
(5) so this is too for the geometric distance error (
s dg ) and for zenithal angle error ( s l ).

Figure 1. Deflection measurement: reference systems, observable and objective measurands.

3.3 Surveying techniques to measure rotation where s xP , s zP are estimated errors of the abscise
The determination of the rotations of each sec- and applicate (z-coordinate) or height, respec-
tion of the beam between two different load tively; and d’ is the distance between reflector
states (i), through surveying measurements, re- centres.
quire to measure the position of two reflectors,
situated in the superior and inferior flanges of the
4 NUMERICAL SIMULATIONS
beam, in both load states.
4.1 Cross-Entropy algorithm for estimation of
Applying the same principles as in Section 3.2,
structural stiffness from static measurements
the estimated error of the rotations () is:
Given the distribution of flexural stiffness (modu-

s r = 2 × arctan
(
max s xP ,s zP ) (6)
lus of elasticity x inertia) of a beam of known
d¢ / 2 boundary conditions and fixed static loading con-
ditions, its deflection profile is unique. In the in-
verse problem consisting of a limited number of consist of a uniformly distributed load (equiva-
available displacements (possibly corrupted), a lent to the self-weight of the bridge deck) and two
number of stiffness distributions may be able to concentrated loads of 10 t each separated by 5 m
reproduce these displacements. The CE algorithm and located symmetrically with respect to mid-
relies on a FE model to establish the global stiff- span (simulating 2-axle truck). Vertical deflec-
ness matrix of the beam, the latter being function tions are simulated at every node of the FE mod-
of the flexural stiffness of each discretized FE and el. Displacements from points within 2 m from
its length. The CE method proposes an iterative the supports are not used as inputs to the algo-
procedure where a number of FE trial beams rithm. These displacements near supports are ig-
(TBs) are generated based on stiffness values that nored because they are small and of order of
are randomly sampled from a predefined statisti- magnitude similar to the expected noise, increas-
cal distribution (i.e., normal) of mean and stand- ing computational without necessarily improving
ard deviation updated in each iteration. Thus, a the results.
different set of TBs with different stiffness pro-
While HDS allows measuring a point every 1 mm
files and, consequently, different deflection pro-
(100 elements/m), such a high number of ele-
files, is obtained at each iteration of the algo-
ments increases the computational time in ex-
rithm. In particular, a total of 1000 TBs are
cess, and a measurement density of 50 displace-
generated in each iteration in this paper. The de-
ments/m is adopted here for the HDS equipment.
flection of each TB is compared against that of the
A measurement density of 10 displacements/m is
true beam and those TBs performing best (lowest
assumed to simulate data gathered by TAC
mean square error of the deflections at measured
equipment. Two discretization levels are consid-
points) are selected to update statistical distribu-
ered for the FE beam, 10 and 50 elements/m, to
tions associated to each discretized beam ele-
facilitate the implementation of the algorithm
ment for the next iteration. In this paper, the best
when providing TAC and HDS measurements re-
10% of all TBs are retained. The process is re-
spectively. In addition, the same stiffness value is
peated until the rate of change of error of the ob-
assigned to all the elements contained within one
jective function (sum of the squared differences
meter to speed up calculations, instead of consid-
between target/measured deflections and pre-
ering an individualized value for each discretized
dicted deflections using FE model for each obser-
element.
vation point) falls below 0.01% over 10 itera-
tions. Then, injection (artificial widening of the Four case scenarios are considered for both TAC
statistical distribution) is applied to prevent the and HDS data: (i) noise-free measurements from
algorithm falling into local minima. This injection healthy beam, (ii) noisy measurements from
is repeated for three convergences of the algo- healthy beam, (iii) noise-free measurements from
rithm. Values for the initial statistical distribu- damaged beam, and (iv) noisy measurements
tions of flexural stiffness for each discretized FE from damaged beam. Damage is modelled as a
are random, although the mean is selected to be 40% stiffness reduction between 12 m and 13 m.
within expected theoretical values and the stand- As such a sudden drop in stiffness is not likely in
ard deviation is set at 20% of the mean value. a real situation, a 10% stiffness reduction is ap-
Further details can be found in Walsh & González plied in adjacent meters to smooth the stiffness
(2009). profile.
4.2 Theoretical scenarios for testing 4.3 Noise model
A simply supported beam, of span length 21 m, is The nominal accuracies of the selected instru-
used to test the performance of the algorithm. A ments (table 1) that have any influence in the de-
flexural stiffness of 3450 MN·m2, typical of a 21 m flection errors are:
bridge, is adopted. In all cases, the applied loads
Table 1. Nominal accuracies for Leica TCRP1201 ta- sui, TAC, 95 % (mm)
d HZ (horizontal distance in m)
15,00 21,00 30,00
chymeter and Leica ScanStation2 laser-scanner. 0,00 0,289 0,405 0,578

h (height in m)
3,00 1,588 1,202 0,990
TAC HDS 6,00 2,973 2,238 1,687
(Leica TCRP1201) (Leica ScanStation2) 9,00 4,114 3,185 2,394
distance accuracy 1 2
2 mm + 2 ppm · d g 4 mm 12,00 5,000 4,004 3,061
( sdg) 15,00 5,669 4,692 3,669
angular accuracy 1
horizontal distance (m)
1" 60 mrad = 12,38" sui, HDS, 95 % (mm)
(s ang) 15,00 21,00 30,00
compensator accuracy 0,00 3,529 4,941 7,058
0.5" 1"
( scomp)

height (m)
3,00 4,681 5,416 7,229
All the accuracies correspond to one sigma (k = 1). 6,00 6,809 6,555 7,699
1
According with ISO 17123-3 and ISO17123-4. 9,00 8,805 7,910 8,374
2
Out of rules and at a 0 - 50 m range. 12,00 10,411 9,216 9,151
15,00 11,635 10,367 9,949
Maximum deflection errors, estimated in the pre-
analysis (see Section 3.1), at a 95% of confidence
level, are shown in table 2 for TAC and HDS fol- 5 THEORETICAL TESTING
lowing application of Equations 3, 4, 5. They are
5.1 Noise-free simulations
given for combinations of horizontal distances
and heights of the target points with respect to When deflection measurements are noise-free,
the measurement instrument. For the theoretical the CE algorithm is reliable in both no damage
testing in the section that follows, a 6 m high and damage situations for the load and structural
bridge is assumed (relevant values are in bold in model under investigation (Section 4.2). The es-
table 2). It is also assumed that the station is lo- timated stiffness profiles are consistent with the
cated at an optimal distance that minimizes the expected targets and able to capture the localised
quadratic sum of the deflection errors along the loss of stiffness (Fig. 2). The estimated peak loss
beam. Finally, the theoretical deflections that re- of stiffness is relatively smaller and spread over a
sult from the beam model in Section 4.2 are con- slightly wider length than the target value. How-
taminated using an additive noise model (i.e., ever, the overall loss of stiffness and critical loca-
adding a value randomly sampled from a statisti- tion where it occurs, are correctly identified.
cal distribution modelling noise to the true de-
The difference in stiffness profiles shown in Fig-
flection). The noise model is given by a normal
ure 2 for different runs of the CE algorithm (i.e.,
distribution of mean zero and a standard devia-
test1, test2, test3, test4 and test5 in the figure) is
tion that depends on the geometrical distance
a result of the random nature of the sampling
from each measurement point to the station.
through each iteration (Section 4.1). The uncer-
tainty associated to the nature of the algorithm
can be reduced by using an average of a number
Table 2. Deflection errors for Leica TCRP1201 ta- of CE tests as shown in the figure. Therefore, the
chymeter and Leica ScanStation2 laser-scanner in
‘average’ curve is the closest curve to the target
many probable observation conditions.
profile in the figure. Figures 2a, b are obtained us-
ing measurements every 10 cm along beam and
Figures 2c, d are obtained using measurements
every 2 cm (excluding 2 m at each end support).
 to the CE algorithm to check their impact on the
estimated stiffness (Fig. 3). A small improvement
is noticeable in the results of the healthy beam,
i.e., the profile in Figure 3a being smoother and
closer to the target value than in Figure 2a.

Figure 3. Estimated stiffness profiles using 10 meas-

urements/m including displacements and rotations:
(a) healthy beam, (b) damaged beam.

5.2 Noisy simulations

Noise is taking into account following the model-
ling procedure of Section 4.3 and it depends on
Figure 2. Target and estimated stiffness profiles using the location and precision of the equipment being
noise-free deflection measurements: (a) 10 measure- simulated (TAC or HDS).
ments/m in a healthy beam, (b) 10 measurements/m
in a damaged beam, (c) 50 measurements/m in a
healthy beam, (d) 50 measurements/m in a damaged When simulated TAC measurements are consid-
beam. ered (10 measurements/m), the CE algorithm
generally shows that the stiffness drops close to
damaged elements (between 12 and 13 m) as
An increase in measurement density from 10 shown by Figure 4, but drops can also be found in
(Fig. 2a) to 50 measurements/m (Fig. 2c) im- healthy elements. Therefore it could be said that
proves the estimation of stiffness distribution in a the predictions by the CE algorithm are inconclu-
healthy beam of uniform stiffness (i.e., leading to sive for this particular structure and loading.
a smoother and closer profile to the target). How-
ever, no significant differences appear in the
stiffness estimation of the damaged beam when
using 10 (Fig. 2b) or 50 measurements/m (Fig.
2d). As it has been previously noticed (Walsh &
Gonzalez, 2009; Gonzalez et al. 2013), stiffness
predictions for the elements closest to the sup-
ports are unreliable due to the low sensitivity of
deflection to the stiffness of these areas. Meas-
ured rotations are included as additional inputs
Figure 4. Target and estimated stiffness profiles using ement mechanical properties), loading conditions
simulated TAC measurements from damaged beam.
and set of CE parameters. This noise limit may
clearly be increased for more flexible structures
(or larger loads) leading to higher displacements
Simulated HDS measurements (50 measure-
or decreased for stiffer structures (or smaller
ments/m) leads to higher deviations with respect
loads) leading to smaller displacements.
to the target stiffness profiles than TAC meas-
urements as a consequence of higher noise asso-
ciated to this measuring device. 6 CONCLUSIONS

If simulated noisy rotations are included together The interest of using surveying measurements for
with displacements in the CE calculations, there structural assessment as opposed to other sen-
is not an improvement in the estimated profiles sors that need direct contact with to the structure
with respect to Figure 4. is beyond any doubt. This paper has used simu-
lated deflection and rotation surveying meas-
Finally, several different values of deflection er-
urements to assess the accuracy of a CE algorithm
rors are tested, with the aim of establishing the
in estimating the stiffness profile of a structure.
noise limit acceptable for accurate stiffness pre-
For this purpose, sources of error related to the
diction using CE in a structure of characteristics
equipment (TAC and HDS) used to collect static
and loading similar to the one described in Sec-
measurements have been quantified. For the se-
tion 4.2.
lected structure (21 m simply supported bridge)
A simplified noise model, where all points are at- and loading case (self-weight + 2-axle truck), the
tributed the same maximum measurement error CE algorithm has shown to be reliable when ap-
is adopted. Figure 5 illustrates the ability of the plied to noise-free data, using measurement den-
CE algorithm to predict damage for various de- sities of 10 (TAC) and 50 (HDS) points per me-
flection errors. The noise limit to identify the ter. However, the corruption of the
stiffness profile with reasonable degree of accu- displacements by noise associated to TAC or HDS
racy varies in a threshold between 0.4 mm and measurements has prevented the accurate esti-
0.8 mm. Beyond 0.8 mm, damage is not unambig- mation of the stiffness profiles by the CE algo-
uously identified. rithm.

The noise limit that would allow reliable stiffness

profiles and to detect damage through the CE al-
gorithm is not achievable under the test condi-
tions in this paper. Further developments in sur-
veying techniques appear to be necessary to
achieve an adequate prediction of structural pa-
rameters using CE and static measurements as
inputs. It should be noted that accuracy of deflec-
tion and rotation measurements might be higher
Figure 5. Target and estimated stiffness profiles (av- in a real case than in the preliminary analysis car-
erage of 5 simulations) using corrupted deflection ried out in this paper. Noise can be reduced
measurements for various deflection errors and 10 through the use of higher precision tachymeters
measurements/m in a damaged beam.
or digital levels, or through the application of so-
phisticated filtering algorithms to the HDS meas-
urements. Therefore, more flexible structures or
It must be noticed that the noise limit previously higher loads will lead to larger displacements
recommended has been determined for a specific
structure (geometry, boundary conditions and el-
where noise will affect stiffness estimations to a González, A., Covián, E., Casero, M. & Cooper, J. 2013.
Experimental Testing of a Cross-Entropy Algorithm
lesser extent. to Detect Damage. 10th International Conference on
Damage Assessment of Structures. Dublin (Ireland),
8-10 July 2013.
Gordon, S.J. & Lichti, D. 2007. Modeling Terrestrial La-
ACKNOWLEDGMENTS ser Scanner Data for Precise Structural Defor-
mation Measurement. Journal of Surveying Engi-
The authors wish to express their gratitude for neering 133-2: 72-80.
Hjelmstad, K. & Shin, S. 1997. Damage detection and
the support received for Scientific Research un- assessment of structures from static response.
der Grant No. BIA2011-26915, from the “Progra- Journal of Engineering Mechanics 123(6): 568-576.
ma Nacional de Proyectos de Investigación Fun- International Organization for Standardization (ISO).
damental, VI Plan Nacional de Investigación 2001. ISO 17123, Optics and optical instruments —
Field procedures for testing geodetic and surveying
Científica, Desarrollo e Innovación Tecnológica instruments. Geneve (Switzerland): ISO copyright
2008-2011” of the Spanish Government towards office.
this investigation. Li, X. Ge, L., Ambikairajah, E., Rizos, C., Tamura, Y. &
Yoshida, A. 2006. Full-scale structural monitoring
using an integrated GPS and accelerometer system.
REFERENCES GPS Solutions 10: 233–247.
Luis, de, J.M. 2009. Contraste en la ejecución de auscul-
Abdo, M. 2011. Parametric study from static response in struc- taciones geodésicas por métodos clásicos y con láser
tural damage detection. Engineering Structures 34: 124- escáner. Tesis doctoral, Departamento de Ingenie-
131. ría Geográfica y Expresión Gráfica, Cantabria Uni-
Bakhtiari-Nejad, F., Rahai, A. & Esfandiari, A. 2005. A struc- versity, Santander (Spain)
tural damage detection method using static noisy data. En-
gineering Structures 27(12): 1784-1793. Marchamalo, M., Galán, D., Sánchez, J.A. & Martínez, R.
Banan, M. R., Banan, M. R., and Hjelmstad, K. D. 1994a. Pa- 2011. La tecnología DGPS en la construcción: con-
rameter estimation of structures from static response. Part I: trol de movimientos en grandes estructuras. In-
computational aspects. Journal of Structural Engineering formes de la Construcción 63-522: 93-102.
120(11): 3243-3258.
Banan, M. R., Banan, M. R., and Hjelmstad, K. D. 1994b. Pa- Mills, J. & Barber, D. 2004. Geomatics Techniques for
rameter estimation of structures from static response. Part Structural Surveying. Journal of Surveying Engi-
II: numerical simulation studies. Journal of Structural En- neering 130-2: 56-64.
gineering 120(11): 3259-3283.
BIMP (Bureau International des Poids et Mesures). Olalla Marañón, C. 2007. Auscultación de laderas. Jor-
2008. Evaluation of measurement data - Guide to nadas técnicas sobre estabilidad de laderas en em-
the expression of uncertainty in measurement. Paris balses, Confederación Hidrográfica del Ebro, Zara-
(France): Joint Committee for Guides in Metrology. goza (Spain) June 2007.

BIMP (Bureau International des Poids et Mesures). Psimoulis, P.A., Pytharouli, P. & Stiros, S. 2006. Exper-
2012. International vocabulary of metrology – Basic imental monitoring of oscillations of major flexible
and general concepts and associated terms (VIM). structures using GPS and RTS. 3rd IAG / 12th FIG
3rd edition. Paris (France): Joint Committee for Symposium. Baden (Austria) 22-24 May 2006.
Guides in Metrology.
Sanayei, M. & Scampoli, S. 1991. Structural element
Boehler, W. & Marbs, A. 2004. Investigating Laser stiffness identification from static test data. Journal
Scanner Accuracy. i3mainz, Institute for Spatial In- of Engineering Mechanics 117(5): 1021-1036.
formation and Surveying Technology, FH Mainz,
University of Applied Sciences, Mainz, Germany. On Walsh, B.J. & Gonzalez, A. 2009. Assessment of the
line: http://scanning.fh-mainz.de. condition of a beam using a static loading test. Key
Engineering Materials 413-414: 269-276.
Caddemi, S. & Morassi, A. 2007. Crack detection in elastic
beams by static measurements. International Journal of Wang, X., Hu, N., Hisao Fukumaga & Yao, Z.H. 2001.
Solids and Structures 44: 5301-5315. Structural damage identification using static test
Cheng Penggen, Shi Wenzhong & Zheng Wanxing. 2002. data and changes in frequencies. Engineering Struc-
Large Structure Health Dynamic Monitoring Using GPS tures 23: 610-621.
Technology. Engineering Surveys for Construction Works
and Structural Engineering FIG XXII International Con-
Yan, Y.J., Cheng, L., Wu, Z. Y. & Yam, L.H. 2007. Devel-
gress; Washington, DC (USA), 19-26 April 2002.
opment in vibration-based structural damage de-
Covián, E. & Puente, V. 2013. Fundamentos del ajuste de ob-
tection technique. Mechanical Systems and Signal
servaciones topográficas. Oviedo (Spain): Publication Of-
Processing 21: 2198-2211.
fice of the Oviedo University. Yang, Q. & Sun, B. 2010. Structural damage localiza-
Ghiliani, P. & Wolf, R. 2006. Adjustment computations. tion and quantification using static test data. Struc-
Spatial data analysis. New Jersey (USA): John Wiley tural Health Monitoring 10(4): 381-389.
& Sons.