Upload
others
View
8
Download
0
Embed Size (px)
Citation preview
Quantitative Damage Identification for Prestressed Tendon Anchorage using PZT Interface Technique and Model Updating
Thanh-Canh Huynh1), Ngoc-Loi Dang2), Joo-Young Ryu2)
and *Jeong-Tae Kim3)
1), 2, 3) Department of Ocean Engineering, Pukyong National University, Busan, Korea *) [email protected]
ABSTRACT
As the impedance monitoring technique has been successfully implemented on various types of structures, the quantitative damage detection by the impedance-based technique has become an important issue for complex structural connections. Since the impedance-based technique is not based on any particular physical models, it is very difficult to quantitatively correlate the change in impedance signatures to the change in structural properties. In this study, the severity of damage in tendon anchorage caused by the loss of tendon forces is quantitatively identified by using the PZT interface-based impedance monitoring technique. Firstly, a prototype of PZT interface is designed for the impedance monitoring of tendon anchorage. Next, impedance signatures are experimentally measured from a lab-scaled tendon-anchorage system under different prestress-force levels. Finally, damage severities of the tendon anchorage caused by the loss of prestress-force are quantitatively identified from a phase-by-phase model updating process, in which changes in the measured impedance signatures are correlated to changes in structural properties. 1. INTRODUCTION
Many researchers have studied the impedance-based damage detection in
structural systems during the past decade. The primary advantages of the impedance-based technique include (1) the technique is not model-based, so it can be easily applied to complex structures; (2) the technique uses small, cheap, lightweight, fast response, and non-instructive sensors; (3) the technique itself is very sensitive to small damage-size; and (4) the technique can be integrated with a wireless impedance sensing network to perform auto-monitoring. By taking the above advantages, the impedance-based technique has been applied for structural health monitoring of tendon anchorage systems (Liang et al. 1994, Kim et al. 2010, Nguyen and Kim 2012, Huynh
1) PhD 2) Graduate Student 3) Professor
and Kim 2014, Huynh et al. 2014, 2016). As the impedance monitoring technique has been successfully implemented on the tendon anchorage, the quantitative damage detection by the impedance-based technique has become an important issue. Most of experimental studies used the impedance technique for qualitative health monitoring (Chaudhry et al. 1996, Park G. et al. 2001, Kim et al. 2010, Park et al. 2015, Huynh and Kim 2016, Nguyen et al. 2017). In these applications, the impedance signatures have been qualitatively classified for the alert of damage occurrence due to the target structure’s complexity and the difficulty in high-frequency modelling.
Up-to-date, few researchers have reported on the quantitative health monitoring in structural joints using the impedance-based method. Ritdumrongkul et al. (2004) employed the spectral element method to model the PZT-bolted joint and applied a mathematical technique to quantitatively estimate structural parameters of the joint. In their study, the considered frequency band was relatively low, less than 2.5 kHz, as compared to the ones commonly used in reality. Lopes et al. (2000) and Min et al. (2012) combined the impedance-based technique with the ANN (artificial neural network) to estimate damage types and severities in joints. However, a large amount of the impedance records should be conducted on the target structure under both intact and damage states that could limit the application of their methods to realistic structures.
In this study, the severity of damage in tendon anchorage caused by the loss of tendon forces is quantitatively identified by using the PZT interface-based impedance monitoring and model updating technique. Firstly, a prototype of PZT interface is designed for the impedance monitoring of tendon anchorage. Next, impedance signatures are experimentally measured from a lab-scaled tendon-anchorage system under different prestress-force levels. Finally, damage severities of the tendon anchorage caused by the loss of prestress-force are quantitatively identified from a phase-by-phase model updating process, in which changes in the measured impedance signatures are correlated to changes in structural properties. 2. PZT INTERFACE-BASED IMPEDANCE MONITORING METHOD
2.1 Prototype of Mountable PZT Interface To predetermine effective frequency bands in impedance-based damage
detection, a mountable PZT interface device was developed by Huynh and Kim (2014). The interface prototype is a beam-like structure with a PZT patch at the center, see Fig. 1(a). The interface device has a flexible beam section in the middle and the fixed-fixed boundary by two outside contact bodies. The flexible section, where the PZT sensor is installed, is intentionally designed to allow flexural vibrations of the interface according to piezoelectric deformations of the PZT. The two contact bodies of the PZT interface allows it to be easily attached and detached from the target structure.
As shown in Fig. 1(b), the PZT interface is surface-mounted on the host structure to monitor the changes in its structural properties which can be induced by any structural damage. Under the excitation of the PZT sensor, the PZT interface interacts with the mother structure via the coupling vibrations. If the mother structure is damaged, the coupling dynamic responses could be altered, making the variation of the EM
impedan
2.2Se
the PZTresonandynamicinterfaceinterfacethe intemodulusdamping2015).
A interfacewas appharmoniof the PZnumericpeak fre
Mothe first them, M
nce respon
(a) PrototypFi
2 Local Dyensitive freT interface nt peaks ofc charactere device we’s materiaerface. Thes E = 70.0g loss fact
(a) Geo
FE model e was discplied, as ic excitatioZT patch,
cal impedaequencies wodal analyfour mode
Mode 1 is
nses obtain
pe of PZT inig. 1 Desig
ynamic Chaequency ra
for whichf the imperistics of th
was selecteal. A PZT pe propertie0 GPa, Potor = 0.0
ometric sizeFi
of the PZTcretized by shown in
on voltage and the bo
ance respowere ident
ysis was nue shapes out-of-plan
ned from th
nterface gn of PZT i
aracteristicanges of im its dynam
edance sighe PZT inted, as showpatch sizedes of the
oisson's rat02. The P
es g. 2 Geom
T interface608 solid Fig. 2(b).with an am
ottom one onses of thified at 19.umerically representinne longitud
he interfac
(bnterface d
cs of PZT Impedance mic characgnatures. Tterface. Fown in Fig. d 15x15x0aluminum
tio = 0.3PZT patch
metries of P
e was esta
elements . To acqumplitude owas set ashe PZT in.18 kHz anperformedng the modinal flexur
e.
) PZT interfevice (Huy
Interfacesignatures
cteristics isThus, it is or that purp
2(a). Alu0.51 mm3 wm interface33, mass dh was add
(b) MeshingPZT interfa
blished byand the fix
uire the Ef 1 V was s the grounterface, in
nd 38.44 kHd for the PZodal vibratiral motion
face-host stynh and Ki
s can be cs closely aneeded to
pose, an exuminum wawas installee are specdensity =ed by PZT
g and bounce (mm)
y COMSOLxed-fixed bM impedasimulated
nd electrodn which theHz, respecZT interfacons of theand Mode
tructure intem 2014)
controlled associated o analyze xample of as selecteed at the mcified as: = 2700 kgT-5A (Huy
dary condit
L Multiphysboundary cance respo to the topde. Figure e first and
ctively. ce. Figure e interfacee 4 is out-
eraction
by using with the the local the PZT d for the
middle of Young's /m3, and
ynh et al.
tions
sics. The condition onses, a p surface 3 shows
d second
4 shows . Among -of-plane
lateral finterfacekHz (Mocharacte
Fig.
3. IMPE
3.As
test stru
flexural moe identical ode 1) aneristics of t
. 4 Modal s
DANCE M
1 Test Sets shown inucture. A 7
otion. By to the two
nd 37.923 the PZT int
Fig. 3 Num
shapes of P
MONITORIN
tup of Lab-n Fig. 5(a), 7-wire mon
comparingo impedankHz (Mod
terface.
merical im
PZT interfasignatures
NG OF LA
-scaled Caa lab-scal
no tendon
g Fig. 3 wnce peaks de 4) that
pedance re
ace identics (Huynh e
AB-SCALE
ble Structule tendon a(15.2 mm
with Fig. 4were num
t mainly r
esponses
cal to resonet al. 2015)
E TENDON
ure anchoragem) was ins
4, two modmerically anrepresent
of PZT inte
nant peaks)
N ANCHOR
e system wstalled in l
dal shapenalysed asthe local
erface
s of impeda
RAGE
was selectelength of 6
es of the s 19.016 vibration
ance
ed as the 6.4 m by
steel anThe destendon-aharmonimeasureintroducone ancforce T140.2 kN5 measu
3.2Fig
under vfrequenctendon-fwas grawere gra35.60 kHthe inte
nchorage hsigned PZTanchorageic voltage ed by an
ce the preschorage to 1 = 48.1 k, T3 = 29.4urements w
(a
Fig. 6 Imp
2 Impedangure 6 sho
various precy bands, force scen
adually redadually decHz (Peak rface-anch
heads (45T interface
e connectiowith 1V-aimpedanc
stress forcmeasure
N was intr4 kN and Twere condu
a) Tendon aFig. 5 Te
pedance re
nce Responows real imestress for
14-20 kHznarios and duced fromcreased fro2). By not
horage sys
5 mm) on e was mouon, as shmplitude,
ce analysece into the
the actualroduced. NT4 = 20.6 kucted repe
anchorage est setup fo
esponses o
nses undempedance rces. Two z (Peak 1the variati
m 48.1 kN om 17.18 kticing that stem was
bearing punted on thown in Fand the im
er, HIOKI cable sys
l cable forNext, the pkN to simueatedly. Th
(b) Inor PZT inte
of PZT inte
er Prestresssignaturesresonant ) and 34-4ion of the (T1) to 20
kHz to 16.6the changcaused b
plates (100the surfaceig. 5(b). T
mpedance 3532. A
stem. And,ce. As the
prestress folate the dae tempera
nterface on erface-anc
erface unde
s Force Chs in 10-100frequencie
40 kHz (Ppeak frequ0.6 kN (T468 kHz (Pe
ge in coupby the cha
0x100x10 me of the beThe PZT signals instressing a load ce
e healthy sorce was d
amage casture was k
bearing plahorage sys
er various
hanges 0 kHz fromes were ideak 2). Tauencies. A4), the twoeak 1) andled impedaange in te
mm3) at twearing plawas excit
n 10-100 kjack was
ell was insstate, the pdecreased
ses. For eakept as 23-
ate stem
prestress f
m the PZT dentical to able 1 out
As the tendo peak freqd from 36.7ance respendon forc
wo ends. te of the ed by a
kHz were used to
stalled at prestress d to T2 = ach case, -24 oC.
forces
interface the two
lines the don force quencies 78 kHz to onses of ces, it is
reasonable to confirm the relationship between the prestress force and the system stiffness. That is, the prestress force-loss caused the reduction in the modal stiffness of the coupled interface-anchorage system.
Table 1 Peak frequencies of PZT interface under various tendon-force levels
Case Tendon Force
Level (kN) Peak 1 Peak 2
Freq. (kHz) Var. (%) Freq. (kHz) Var. (%)T1 48.1 17.18 0 36.78 0 T2 40.2 17.04 -0.81 36.62 -0.44 T3 29.4 17.00 -1.05 36.50 -0.76 T4 20.6 16.68 -2.91 35.60 -3.21
4. QUANTITATIVE DAMAGE IDENTIFICATION OF TENDON ANCHORAGE Damage severities induced by the prsestress-loss in the tendon anchorage
should be quantified to estimate the variation of structural performance with regards to its designed state. This goal is achieved by the following steps: (1) a FE model of the PZT interface-tendon anchorage system is simulated; (2) a phase-by-phase model updating process is designed to fine-tune the FE model; and (3) the model updating is performed to correlate the change in impedance signatures to the change in structural properties, both of which are induced by the variation of prestress forces.
4.1 FE Analysis of Tendon Anchorage with PZT Interface Figure 7 shows a FE model of the tested tendon anchorage with the PZT
interface which was established by using COMSOL Multiphysics. The dimension of the FE model was consistent with the actual size of the target structure. The presence of the tendon force can be equivalently treated as the contact stiffness of the bearing plate (Nguyen and Kim 2012, Huynh et al. 2015). The loss of the prestress force can be simulated by the reduction in the contact stiffness. Hence, an equivalent contact spring system kx, ky and kz was simulated for the test structure (see Fig. 7(a)). The loss factor of the contact damping is assumed as 0.02. The FE model was meshed by 1463 solid elements.
The bonding layers between the PZT patch and the interface were not considered for the simplification. The material properties of the aluminum interface were input to the FE model, as described in the previous section. The steel tendon-anchorage has Young's modulus E = 200 GPa, Poisson's ratio = 0.3, mass density = 7850 kg/m3, and damping loss factor = 0.02. The material properties of the PZT patch can be found in Huynh et al. (2015). The initial contact stiffness parameters were selected as: kx = ky = 0 and kz = 106 N/m/m2. A harmonic voltage of 1V-amplitude was simulated to the top of PZT, and the bottom one was set as the ground electrode.
Figure 8 shows real impedance signatures which were generated from the initial FE model. According to the figure, two resonant impedance peaks were identified at f1 = 18.16 kHz (Peak 1) and f2 = 36.96 kHz (Peak 2). As compared to the experimental results shown in Fig. 6, the two calculated impedance peaks were respectively consistent with the two experimental ones, though the peak frequencies and peak
magnituwhile tha
Fig. 4.2As
would bthe interthe locaindicatorscaled tthe corre
Figparticulaparametupdated(2004), tas compperformethis mo
des were at of Peak
(a) FE mo
8 Numeric
2 Phase-bs describee unchangrface devical dynamicr for the ctendon-ancesponding gure 9 sharly designter having
d first (Ho ethe variatiopared to thed for the del updat
quite differ2 is the lat
odel Fig. 7 FE
cal impeda
y-phase Med previousged unlessce are altec charactecontact stifchorage ca FE modelhows the ned for the
higher seet al. 2012on of lateraat of axial phase kz
ing proces
rent. The mteral flexur
model of t
ance signat
Model Updasly, the lo
s the structered. By aseristics of ffness conan be iden are well-mphase-by
e tendon aensitivity to2, Huynh eal contact scontact stifirst, and ss needs
motion of Pral mode (s
(b) Interfactendon-anc
tures of ini
ating Algorocal dynamtural paramssuming ththe interfa
ndition. Thntified whematched wy-phase manchorageo the strucet al. 2016)stiffness (iiffness (i.enext was tto be pe
Peak 1 is tsee Fig. 7(
ce’s two mochorage wi
itial FE mo
rithm mic responmeters of the interfacace-anchorat is, the
en the impewith the expmodel upde system (ctural chan). As repo.e., kx and ., kz). Therthe phase
erformed i
he longitud(b)).
odal responsth PZT inte
odel of tend
nses of ththe tendone itself remrage can contact st
edance peperimental ating algo
(Huynh annge would orted by Rky) has rel
refore, the kx. It sh
n combina
dinal flexu
ses erface
don-ancho
he interfacen-anchoragmains undbe utilizedtiffness of eaks obtainresults. orithm wh
nd Kim 20be select
Ritdumronglatively smmodel updould be noation with
ral mode
rage
e device ge and/or amaged, d as the the lab-
ned from
hich was 17). The ed to be kul et al.
mall effect date was oted that
the FE
modelling at each iteration step.
Fig. 9 Model updating algorithm for contact stiffness identification
4.3 Identified Contact Stiffness of Tendon-Anchorage The model updating process was carried out for the tendon anchorage using the
impedance signatures measured from the PZT interface. At first, the model updating was performed for the tendon force T1 = 48.1 kN. For the case T1, three phases were performed to match the numerical impedance peaks (Peaks 1 and 2) to the experimental results. Figure 10 shows convergences of two peak frequencies of impedance signatures during model updating for the case T1. In Phase 1, the contact stiffness kz was first updated. Parameter kz was went up from kz = 106 N/m/m2 to kz = 4x1012 N/m/m2 while kx and ky were remained as zero. After Phase 1, the frequency error of Peak 1 was converged from 5.7 % to 0.58 % while that of Peak 2 was converged from 0.58 % to 0.33 %. In Phase 2, the contact stiffness kx and ky were updated. The contact stiffness was gradually increased from kx = ky = 0 to kx = ky = 2.6x1012 N/m/m2 while kz remained unchanged. After Phase 2, the frequency error of Peak 1 was decreased from 0.58 % to 0.12 % while that of Peak 2 remained unchanged as 0.33 %. In Phase 3, kz was again updated, and the contact stiffness was changed from kz = 4x1012 N/m/m2 to kz = 5.15x1012 N/m/m2 while kx and ky were remained unchanged. After Phase 3, the frequency error of Peak 1 was converged from 0.58 % to 0 % while that of Peak 2 was reduced from 0.33 % to 0.16 %.
From the model updating for T1 = 48.1 kN, it is observed that the two peak frequencies of the selected impedance signatures were changed in their frequencies during the model updating (see Fig. 10). The frequency differences of Peak 1 and Peak 2 were decreased as the iteration C24 was reached (see Fig. 10). The stiffness parameters were identified as the converged values when the peak frequencies of the updated FE model became identical to the experimental peak frequencies.
The same updating process was repeated for all tendon force cases T2-T4, from
Measure Experimental Local Dynamics Measure experimental impedance: Zexp
Identify peak frequency fiexp (i = 1, m)
Establish initial FE model Assumption: kx = ky ≤ kz
Select initial contact stiffness: kxo, kyo, kzo
Perform model update phase-by-phase
Check{} 0
Identified the contact stiffness
Compute resonant frequency difference
YesNo
Select a model update phase (kj)
Analyze numerical impedance response: Znum
Identify resonant frequency: finum (i = 1, m)
Select an increment
of contact stiffness
k j= kjo + kjChange phase ?
YesNo
2
exp1
i
numi
i f
f
which thas showthe modthat theresults a
Fig. 10 C
Fig. 1
he stiffnesswn in Fig. 1del updatine numericaat the same
Convergen
(a) Te
(c) T11 Impeda
s paramete11. In the g were co
al impedane frequenc
nces of two
ndon force
Tendon forcance signat
ers correspfigure, thempared tonce respocy range an
o impedanc
T1
ce T3 tures of up
ponding toe numericao the measnses werend also ide
ce peaks d
pdated FE
o the four tal impedansured expee well-matentical patt
during mod
(b
(d)models for
tendon forcce signatu
erimental otched witherns.
del updatin
) Tendon fo
) Tendon for four tendo
ces were iures identifones. It is oh the expe
ng for T1 =
orce T2
orce T4 on forces T
identified fied from observed erimental
48.1 kN
T1-T4
4.4Re
kz) and examinerelationsrelationspredicte
As
structuraestimatiobetweenusing ththese remeasureprestres
4 Quantitaelationshipthe peak f
ed as shoships sincships, the d for meas
(a) IdentifiFig.
Fig. 13 R
s the finaal contacton of the tn the tendohe exponenelationshiped impedassed tend
ative Identifps betweenfrequencieown in Fce they p
contact ssured peak
ied kx and k 12 Identif
Relationsh
l step, thet stiffnesstendon-ancon forces antial functips, the vaance signaon ancho
fication of n the idents (i.e., Peaig. 12. Porovided thstiffness pk frequenci
ky ied contac
hip between
e relationss is conschorage coand the coons, which
alue of theatures. It orage can
Tendon Antified contaak 1 and Power funche good parametersies of impe
ct stiffness
n contact s
ship betwestructed foonnection, ontact stiffnh providede prestresis observe
n be qua
nchorage Dact stiffnesPeak 2) ofctions werfits for th
s of the tedance sig
(versus pea
stiffness an
een the por the quas shown
ness kx, ky
d the best ss force ced that theantitatively
Damage s parametf impedancre used the data. tendon-ancnatures.
(b) Identifiedak frequen
nd prestres
prestress-louantitative
in Fig. 13y and kz wfits for theould be pe severity
y estimate
ters (i.e., kce signatuto establis From uschorage c
d kz ncies
ss forces
oss force damage
3. The relatwere estable data. Fropredicted f
of damaged by m
kx, ky and res were sh these sing the could be
and the severity
tionships ished by
om using from the ge in the easuring
impedance signatures and analyzing equivalent structural parameters corresponding to the damaged state. It is also noted that the relationship between the identified contact stiffness and the prestress force may also be applicable for similar cable anchorage systems. 5. CONCLUSIONS
In this study, the severity of damage in a lab-scaled tendon anchorage caused by the loss of tendon forces was quantitatively identified by using the PZT interface-based impedance monitoring technique. The change in impedance signatures was correlated to the change in tendon forces by the phase-by-phase model updating process. Then the relationship between the prestress force and the contact stiffness was identified for the tendon-anchorage system. By using the relationship, the prestress force could be estimated from the measured impedance signatures. From the findings, it is concluded that the proposed approach has the potential to quantitatively estimate the damage severity in the tendon anchorage which is caused by the tendon force-loss. ACKNOWLEDGMENT
This work was supported by Basic Science Research Program of National Research Foundation of Korea (NRF 2016R1A2B4015087). The post-doctoral researcher and graduate students were also supported by the Brain Korea 21 Plus program of Korean Government. REFERENCES Chaudhry, Z., Lalande, F., Ganino, A., and Rogers, C. (1996), "Monitoring the integrity
of composite patch structural repair via piezoelectric actuators/sensors", AIAA-1996-1074-CP.
Efunda, Inc. (2010), http://www.efunda.com Ho, D.D., Kim, J.T., Stubbs, N., and Park, W.S. (2012), “Prestress-force estimation in
PSC girder using modal parameters and system identification”, Advances in Structural Engineering, 16(6), 997-1012.
Huynh, T.C., and Kim, J.T. (2014), “Impedance-based cable force monitoring in tendon-anchorage using portable PZT-interface technique”, Mathematical Problems in Engineering, 2014, 1-11.
Huynh, T.C., and Kim, J.T. (2016), “Compensation of temperature effect on impedance responses of PZT interface for prestress-loss monitoring in PSC girders”, Smart Structures and Systems, 17(6), 881-901.
Huynh, T.C., and Kim, T.C. (2017), “Quantitative damage identification in tendon anchorage via PZT interface-based impedance monitoring technique”, Smart Structures and System, 20(2).
Huynh, T.C., Lee, K.S., and Kim, J.T. (2015), “Local dynamic characteristics of PZT impedance interface on tendon-anchorage under prestress force variation”, Smart Structures and Systems, 15(2), 375-393.
Huynh, T.C., Nguyen, T.C., Choi, S.H., and Kim, J.T. (2016), “Impedance monitoring at tendon-anchorage via mountable PZT interface and temperature-effect compensation”, Proceedings of SPIE, 9799, 1-6.
Huynh, T.C., Park, J.H., and Kim, J.T. (2016), “Structural identification of cable-stayed bridge under back-to-back typhoons by wireless vibration monitoring”, Measurement, 88, 385-401.
Huynh, T.C., Park, Y.H., Park, J.H. and Kim, J.T. (2014), “Feasibility verification of mountable PZT-interface for impedance monitoring in tendon-anchorage”, Shock Vibration, 2015, 1-11.
Kim, J.T., and Huynh, T.C. (2016), “Impedance-based monitoring of tendon-anchorage via PZT material-aluminum interface technique”, Journal of the Korean Society for Advanced Composite Structures, 7(4), 91-100.
Kim, J.T., Park, J.H., Hong, D.S., and Park, W.S. (2010), “Hybrid health monitoring of prestressed concrete girder bridges by sequential vibration-impedance approaches”, Engineering Structures, 32, 115-128.
Liang, C., Sun, F.P. and Rogers, C.A. (1994), “Coupled electro-mechanical analysis of adaptive material - Determination of the actuator power consumption and system energy transfer”, Journal of Intelligent Material Systems and Structures, 5, 12-20.
Lopes, V., Park, G., Cudney, H.H., Inman, D.J. (2000), “Impedance-based structural health monitoring with artificial neural networks”, Journal of Intelligent and Material Systems and Structures, 11, 206-214.
Min, J., Park, S., Yun, C.B., Lee, C.G., and Lee, G. (2012), “Impedance-based structural health monitoring incorporating neural network technique for identification of damage type and severity’, Engineering Structures, 39, 210-220.
Nguyen, K. D., and Kim, J.T. (2012), “Smart PZT-interface for wireless impedance-based prestress-loss monitoring in tendon-anchorage connection”, Smart Structures and Systems, 9(6), 489-504.
Nguyen, T.C., Huynh, T.C., Yi, J.H., and Kim, J.T. (2017), “Hybrid bolt-loosening detection in wind turbine tower structures by vibration and impedance responses”, Wind and Structures An International Journal, 24(4), 385-403.
Park, G., Cudney, H. H., and Inman, D. J. (2001), “Feasibility of using impedance-based damage assessment for pipeline structures”, Earthquake Engineering and Structural Dynamics, 30, 1463-1474.
Park, J.H., Huynh, T.C., and Kim, J.T. (2015), “Temperature effect on wireless impedance monitoring in tendon anchorage of prestressed concrete girder”, Smart Structures and Systems, 15(4), 1159-1175.
Ritdumrongkul, S., Abe, M., Fujino, Y. and Miyashita, T. (2004), “Quantitative health monitoring of bolted joints using a piezoceramic actuator-sensor”, Smart Materials and Structures, 13, 20-29.