Seismic testing of large scale structures with dissipator devices
Abstract
Seismic testing of large-size models is nowadays possible thanks to the pseudodynamic (PsD) method, which resembles a quasi-static method with the application of slowly varying forces to the test structure. Low testing velocity may change the behaviour of the materials with respect to their behaviour at real velocity due to strain rate effect (SRE). The SRE during a PsD test is not relevant for the classical construction materials. However, it can be very important for some new materials, such as rubber or silicon, mostly used in isolators or energy dissipative devices. Silicone fluid viscous dampers are an example of these dissipative devices, and they will be the subject of this paper. It is possible to extend the PsD method range of application to structures equipped with dissipative devices by introducing a correction technique (CT) developed by the European Laboratory for Structural Assessment (European Commission) in Ispra, Italy. This technique consists of an on-line linear modification of the restoring measured forces. In this paper, two new procedures of CT are proposed. The first procedure consists of a linear filter where the output is the corrected force and the inputs are the measured forces and displacements. The second procedure is based on neural networks, using the same input but adding the test velocity rate. First, the most appropriate scheme of the algorithms was selected and calibrated. This is based on experimental data obtained by dynamically testing the dissipation devices with a displacement pattern, which was specifically designed for this purpose, at different frequencies and test velocities. Then, the capability of the obtained filter/net is tested using experimental data. Additionally, the most appropriate frequency for the training tests was studied. Outcomes were successful. Both algorithms outperform the previous linear CT, the net-based one being the most accurate and versatile. It was found that the first natural frequency of the unprotected structure is an adequate reference for the training tests.
Full text
11
h
In e na ional Con e ence on Vib a ion P oblems
Z. Dimi o o á e al. (eds.)
Lisbon, Po ugal, 9-12 Sep embe 2013
SEISMIC TESTING OF LARGE SCALE STRUCTURES WITH
DISSIPATOR DEVICES
Bea iz Zapico*
1
, José L. Zapico
2
, F ancisco J. Molina
3
1
ARUP B.V. Ams e dam, The Ne he lands
Bea iz.za[email p o ec ed]
2
Depa men o Cons uc ion and Manu ac u ing Enginee ing, Uni e si y o O iedo, Gijon, Spain
jza[email p o ec ed]
3
ELSA Lab, JRC, Eu opean Commission, Isp a, I aly
ancisc[email p o ec ed].eu opa.eu
Keywo ds: seismic es ing, pseudo-dynamic me hod, dissipa i e de ices, ull scale models,
s ain a e e ec , Neu al Ne wo ks
Abs ac . Seismic es ing o la ge-size models is nowadays possible hanks o he pseudo-
dynamic (PsD) me hod, which esembles a quasi-s a ic me hod wi h he applica ion o slowly
a ying o ces o he es s uc u e. Low es ing eloci y may change he beha iou o he ma-
e ials wi h espec o hei beha iou a eal eloci y due o s ain a e e ec (SRE). The SRE
du ing a PsD es is no ele an o he classical cons uc ion ma e ials. Howe e , i can be
e y impo an o some new ma e ials, such as ubbe o silicon, mos ly used in isola o s o
ene gy dissipa i e de ices. Silicone luid iscous dampe s a e an example o hese dissipa i e
de ices, and hey will be he subjec o his pape . I is possible o ex end he PsD me hod
ange o applica ion o s uc u es equipped wi h dissipa i e de ices by in oducing a co ec-
ion echnique (CT) de eloped by he Eu opean Labo a o y o S uc u al Assessmen (Eu o-
pean Commission) in Isp a, I aly. This echnique consis s o an on-line linea modi ica ion o
he es o ing measu ed o ces. In his pape , wo new p ocedu es o CT a e p oposed. The
i s p ocedu e consis s o a linea il e whe e he ou pu is he co ec ed o ce and he inpu s
a e he measu ed o ces and displacemen s. The second p ocedu e is based on neu al ne -
wo ks, using he same inpu bu adding he es eloci y a e. Fi s , he mos app op ia e
scheme o he algo i hms was selec ed and calib a ed. This is based on expe imen al da a ob-
ained by dynamically es ing he dissipa ion de ices wi h a displacemen pa e n, which was
speci ically designed o his pu pose, a di e en equencies and es eloci ies. Then, he
capabili y o he ob ained il e /ne is es ed using expe imen al da a. Addi ionally, he mos
app op ia e equency o he aining es s was s udied. Ou comes we e success ul. Bo h al-
go i hms ou pe o m he p e ious linea CT, he ne -based one being he mos accu a e and
e sa ile. I was ound ha he i s na u al equency o he unp o ec ed s uc u e is an ade-
qua e e e ence o he aining es s.
Bea iz Zapico, José L. Zapico, F ancisco J. Molina
2
1 INTRODUCTION
Seismic design is a ela i ely new concep in he his o y o enginee ing. I consis s in di -
e en s a egies o designing ea hquake esis ing buildings o ensu e he heal h, sa e y and
secu i y o he building occupan s and asse s. Some o hose s a egies can be g ouped unde
he name o s uc u al p o ec ion, measu es aken o enhance he esponse o a building unde
a seismic load. S uc u al p o ec ion can be planed du ing he design o he building, o a e i
– o example in his o ical buildings. I can be di ided in o h ee main g oups: base isola ion,
passi e sys ems, and ac i e sys ems. This pape is ocused on he passi e p o ec ion sys ems,
in pa icula on luid iscous silicon based de ices (FVD). They ha e he ad an age o being
easily implemen ed in an exis en s uc u e and no needing an ex e nal powe supply.
No ma i e ega ding seismic design and hence s uc u al p o ec ion has expe ienced a as
e olu ion in he las cen u y. Wi h each new ea hquake new codes ha e been implemen ed
and new philosophies ha e been de eloped, om he addi ion o an equi alen la e al load in
he s a ic s udy o he pe o mance based design. Acco ding o he la e endency, he e-
sponse o he buildings unde an ea hquake canno be assumed bu needs o be obse ed, and
es s ge a i s line posi ion in he c ea ion o new codes o he alida ion o new cons uc i e
me hods.
A well-known me hod o seismic es ing o s uc u es is he shaking able, mainly used o
he es ing o small-scale models which dimensions a e imposed by he inhe en limi a ions o
he able in e ms o maximum es able model mass and size. Seismic es ing o la ge-size
models is nowadays mo e a o dable hanks o he pseudo-dynamic (PsD) me hod in la ge e-
ac ion-wall acili ies, such as he Eu opean Labo a o y o S uc u al Assessmen (ELSA) o
he Join Resea ch Cen e o he Eu opean Commission. In ac , he al e na i e o using la ge
shaking ables, when a ailable, o hea y and all specimens, may also ha e se ious limi a-
ions o accu acy in he applica ion o he inpu and he ep oduc ion o he cha ac e is ics o
he es ed s uc u es. This is mainly due o he dynamic in e ac ion e ec s be ween specimens
and shaking ables [1]. Mo eo e , when p o ec ion sys ems a e applied o he s uc u e, he
scaling o he sys em wi h he equi ed accu acy becomes i ually impossible.
On he o he hand, he PsD es ing me hod has o be ca ied ou a a low eloci y a e, in
o de o ensu e minimum con ol e o s and powe o he ac ua o s. This low es ing eloci y
may change he specimen ma e ials beha iou wi h espec o hei beha iou a eal eloci y.
This al e a ion is called s ain a e e ec (SRE) and may appea whene e he ime scale ac-
o is di e en om one. In ac , i may also appea , in he opposi e sense, in he dynamic es
on he shaking able wi h scaled-down specimens; in his case, due o he dynamic simila i y,
he ime scale ac o can be smalle han one.
In a PsD es , he ime scale is usually be ween 10 and 1000, which means ha ing a es ing
speed om 10 o 1000 imes smalle han he ac ual one. Luckily, his speed a ia ion does
no in oduce a ele an SRE o he classical cons uc ion ma e ials, such as s eel o conc e e.
Howe e , he SRE can be e y impo an o some new ma e ials, such as ubbe o silicon,
mos ly used in isola o s o ene gy dissipa i e de ices, as he be o e men ioned FVD. In spi e
o he inc emen o he es ing eloci y allowed by he implemen a ion o he con inuous PsD
me hod, he SRE on hese ma e ials canno be comple ely neglec ed, as hei appa en s i -
ness can be modi ied in a 10 o a 20%, i no mo e (see Figu e 1).
Bea iz Zapico, José L. Zapico, F ancisco J. Molina
3
Figu e 1: SR E ec : o ce-displacemen cycle a eal speed (black) and o a speed 100 imes smalle (g ey)
I is possible o ex end he PsD ange o applica ion o cases wi h SRE by in oducing a
co ec ion echnique. In he case o buildings p o ec ed by means o FVD, his echnique is
based on he obse a ion o he de ices beha iou in ela ion wi h he s ain a e change. The
p incipal al e a ion consis s o an a enua ion o he measu ed o ce ampli ude, while he
shape o he o ce-de o ma ion cycle emains mos ly in a iable. Consequen ly, he simples
e sion o he co ec ion consis s o a Simple Linea Me hod (SLM), an on-line modi ica ion
o he es o ing measu ed o ces. The co ec ed o ce is ob ained by simply mul iplying he
measu ed o ce by a coe icien , which is a unc ion o he eloci y a e [2].
In he p esen pape , wo new me hods a e p oposed, bo h consis ing in a co ec ion o he
es o ing o ces ca ied ou nume ically in he analy ical pa o he PsD me hod, a e he
measu ing o he o ces and be o e he calcula ion o he nex s ep displacemen .
The i s new me hod is an imp o ed e sion o he SLM, called Fil e Me hod and i con-
sis s in a linea il e whe e he ou pu is he co ec ed o ce and he inpu s a e o be de e -
mined. The second me hod ep esen s a comple ely di e en app oach, based on a i icial
in elligence, using he same inpu a iables as he i s me hod and adding he eloci y a e λ,
enhancing he me hod wi h he possibili y o changing he eloci y a e du ing he es . This
app oach will be e e ed o as Neu al Ne wo k-based Me hod. Bo h me hods will be u he
desc ibed la e .
2 EXPERIMENTAL PART
2.1 Specimen
The specimen used o he es desc ibed in his pape is a 1-s o ey bol ed s eel s uc u e o
one deg ee o eedom wi h o e all dimensions o 3x3x3 me e s. The op o he s uc u e is
o med by a ein o ced conc e e slab. The s uc u e is complemen ed by wo in e ed V b ac-
es, which a e used as a suppo o he FVD (see igu es 1 and 2). This s uc u e has a double
unc ion in his s udy. On he one hand, i was used as cha ac e iza ion bench when de e min-
ing he de ices beha iou . On he o he hand, i was used as es ing specimen du ing he PsD
es s.
The displacemen s o he s uc u e we e measu ed om a no loaded e e ence ame on he
op angles o he s uc u e, using digi al encode s Heidenhain. The ela i e displacemen s o
Bea iz Zapico, José L. Zapico, F ancisco J. Molina
4
he de ices we e measu ed ia ou po en iome ic ansduce s Ge an. Fo ces we e measu ed
using piezoelec ic load cells, one on each pis on (In e ace) and one on each dissipa ion de-
ice (Laumas).
Figu e 2: Specimen layou .
Figu e 3: Seismic p o ec ion sys em FVD.
Bea iz Zapico, José L. Zapico, F ancisco J. Molina
5
Figu e 4: Specimen ins umen a ion.
2.2 Expe imen al cha ac e iza ion
Fi s o all, he cha ac e iza ion es s we e ca ied ou . The inpu o hese expe imen s was
a displacemen his o y, composed o a se ies o semi cycles o sinusoidal shape and di e en
ampli udes, connec ed in such a way ha he discon inui ies in he slope we e a oided. The
design o he inpu signal was based on wo aspec s: he simplici y, because i should be easily
pe o med on di e en ypes o de ices; and uni e sali y, meaning ha he esponse o he
de ices o he inpu should be as comple e as possible, bo h in displacemen s and eloci ies.
The cha ac e iza ion es s we e ca ied ou unde displacemen con ol and a di e en e-
loci y a es (λ=1, 75, 100, 250 and 500) wi h espec i e di e en equencies o he inpu sig-
nal (2, 2.5, 3, 3.5, 4, 4.5 and 5 Hz). The signals ob ained we e p e-p ocessed in o de o
emo e disco dances in bo h ampli ude and phase.
The na u al equency o he s uc u e was a ound 2.5 Hz.
Figu e 5: Cha ac e iza ion displacemen his o y, 2.5Hz.
The esul s o hese expe imen s, measu ed o ces and displacemen s, o med he expe i-
men al da abase. The da a measu ed du ing he es a eal speed (λ=1) ep esen he se o ob-
jec i e alues.
Bea iz Zapico, José L. Zapico, F ancisco J. Molina
6
3 CORRECTION METHODS
PsD es s a e ca ied ou s ep by s ep. A each s ep, he es o ing o ces a e measu ed on
he es ed s uc u e. The cu en alue o hese o ces, along wi h he accele a ion his o y, a e
used o p edic he displacemen o be applied o he s uc u e in he nex s ep. Due o he
p esence o FVD in he s uc u e, SRE will appea , and he measu ed o ces will be di e en
om he o ces measu ed on a eal speed es . The p oposed co ec ion akes place a he ana-
ly ical pa o he PsD me hod, i.e., when he displacemen s o he new s ep a e calcula ed.
The base o bo h co ec ion me hods is he same: he compa ison be ween wo di e en
signals, ob ained om wo es s ca ied ou in he same s uc u e wi h iden ical inpu da a bu
di e en de o ma ion a es. The co ec ion me hods a e uned in o de o minimise he di e -
ence be ween he measu ed and he p edic ed alues o he es o ing o ce.
The p ocess ollowed wi h bo h me hods is e y simila . Fi s , a da abase o expe imen al
da a is gene a ed (cha ac e iza ion es s). The objec i e alues ha e o be p esen in his da a-
base. A e wa ds, a p elimina y s uc u e is selec ed o he algo i hms (inpu s and o de s o
he il e , inpu s and numbe o neu ons o he neu al ne wo k) and hen he algo i hms a e
ained. This s ep is epea ed wi h se e al possible con igu a ions, un il he op imal s uc u e
is eached. Then, he esul an il e /ne is e i ied, using new da a as inpu and checking he
e o (
Mean Squa e E o
o MSE) o he ob ained co ec ion.
3.1 Fil e me hod
The i s new me hod is called he
Fil e Me hod
. I is a linea il e whe e he ou pu is he
co ec ed o ce and he inpu s a e o be de e mined. To do his, he il e has o be ained wi h
expe imen al da a and di e en inpu s and o de s. The analysis o hese es s e eals ha he
op imal esul s o he il e me hod a e eached when using as inpu he measu ed o ces and
displacemen s and hei i s de i a i es in ime, which co esponds wi h o de ze o o he
ou pu a iable and o de one o he inpu a iables o he il e . This solu ion cons i u es a
ade-o be ween accu acy o he p edic ions and ime o p ocessing. I akes o he ollowing
ma hema ical o m:
݂
መ
ሺଵሻ
ሺݐሻ = ܾ
ଵ
ሺఒሻ
∙ ݂
ሺఒሻ
ሺݐሻ+ ܾ
ଶ
ሺఒሻ
∙ ݂
ሺఒሻ
ሺݐ − 1ሻ+ ܾ
ଷ
ሺఒሻ
∙ ݀
ሺఒሻ
ሺݐሻ+ ܾ
ସ
ሺఒሻ
∙ ݀
ሺఒሻ
ሺݐ − 1ሻ
(1)
whe e
݂
መ
ሺଵሻ
ሺݐሻ
is he p edic ed o ce and λ is he speed o he es . Thus, wha needs o be
de ined a e he ou coe icien s ha mul iply he measu ed o ce
and displacemen
d
a he
ime ins an s
and
-1
.
3.2 Neu al Ne wo k-based Me hod
The second me hod is based on a i icial in elligence, using he same inpu a iables as he
i s me hod and adding he eloci y a e λ, enhancing he me hod wi h he possibili y o
changing he eloci y a e du ing he es . Neu al ne wo ks need o be ained wi h expe i-
men al da a also, so ha hei pa ame e s can be adjus ed. This p ocess is ca ied ou in wo
phases. Fi s , he de i a i es o he e o unc ion (di e ence be ween he objec i e alues and
he gene a ed alues) a e ob ained. In a second phase he ne pa ame e s a e op imized, mini-
mizing he e o . Fo hese phases, he
Back-P opaga ion
algo i hm and he
Scaled Conjuga e
G adien
me hod a e used.
Using he cha ac e iza ion es da a as inpu , i was s a ed ha he app op ia e numbe s o
neu ons we e 5 inpu neu ons, 1 in e nal laye o 4 neu ons and 1 ou pu neu on.
Bea iz Zapico, José L. Zapico, F ancisco J. Molina
7
3.3 Applica ion
La e on, a di e en se o da a was used in o de o check he me hods. I consis ed o he
displacemen s and o ces ga he ed du ing a g oup o quasi-s a ic es s, un wi h a seismic dis-
placemen his o y as inpu signal. This se is called
Seismic Applica ion
in his s udy, as i is
he use o he al eady ained algo i hms on seismic da a. The new me hods, applied o he
seismic signals, p o ided good esul s in e ms o NMS e o (3.16% and 2.74% esp.), he
second one being he mos accu a e and e sa ile. Bo h algo i hms ou pe o m he Simple
Linea Me hod in all he analyzed cases, which p esen ed a NMSE o 6.36%.
Figu e 6: Seismic eal speed es measu ed displacemen .
Figu e 7: Fo ce-displacemen cycle, il e me hod. In black, da a om eal speed es s (λ=1), in g ey, da a om
PsD es wi h λ=100, in ed he co ec ed da a o ha es using he il e me hod
Bea iz Zapico, José L. Zapico, F ancisco J. Molina
8
Figu e 8: Fo ce-displacemen cycle, neu al ne wo ks me hod. In black, da a om eal speed es s (λ=1), in g ey,
da a om PsD es wi h λ=100, in ed he co ec ed da a o ha es using he me hod based in neu al ne wo ks.
4 COMPLEMENTARY STUDIES
4.1 Neu al ne wo k based on seismic da a
Using he seismic da a desc ibed in he p e ious pa ag aph, a u he e i ica ion was ol-
lowed on he Neu al Ne wo k-based Me hod. A new ne was ained using he i s hal o
hese da a, while he second pa o he da a was used in he applica ion phase. I is impo an
o ema k ha his p ocedu e is possible only as an academic s udy and no in a eal case, be-
cause i includes he ealiza ion o dynamic seismic es ing on he s udied s uc u e, which is
exac ly wha is ying o be a oid. The applica ion e o ob ained wi h his ne (1.3% app ox.)
is, as expec ed, lowe han he e o p o ided by he p e ious ne . Ne e heless, he di e -
ences a e no high, demons a ing ha he cha ac e iza ion da a base is enough o a p ope
aining and ha he selec ed pa e n p o ides a good cha ac e iza ion o he sys em.
4.2 Selec ion o he cha ac e iza ion inpu da a equency
As has been seen be o e, he e we e 7 di e en cha ac e iza ion inpu signals, iden ical in
shape bu wi h di e en equency. Looking a he esul s o he Seismic Applica ion, i was
ound ha he i s na u al equency 2.5 Hz o he co esponding unp o ec ed (wi hou luid
iscous dampe s) s uc u e cons i u es an app op ia e equency o he p oposed ime-his o y
displacemen pa e n o es he dampe s. This alue can be easily es ima ed in he design
s age h ough nume ical models o ob ained by modal es ing o he ac ual unp o ec ed s uc-
u e.
5 CONCLUSIONS
A me hodology in ended o make PsD es s applicable o s uc u es seismically p o ec ed
by means o luid- iscous dampe s has been de eloped in his pape . I consis s o he on-line
co ec ion o he measu ed es o ing o ces as a unc ion o he o ces and displacemen s ex-
pe imen ed by he dampe s and he speed o he es .
Fo his pu pose, wo di e en p ocedu es ha e been de eloped. The i s one is o mula -
ed as a linea il e algo i hm in which he ou pu s a e he co ec ed o ces, while he meas-
u ed o ces and displacemen s a he dampe s ac as inpu s. The second one uses he same
inpu s and ou pu s, bu adding he es eloci y ac o as a new inpu a iable. In his case,
hese a iables a e unc ionally in e pola ed by means o a neu al ne wo k.
Bea iz Zapico, José L. Zapico, F ancisco J. Molina
9
These algo i hms a e calib a ed on he basis o da a ob ained ia an expe imen al p oce-
du e. This p ocedu e consis s o he dynamical applica ion o a ime-s o y pa e n o dis-
placemen , designed ad hoc o he cha ac e iza ion o he dampe s, and i s cyclic
ep oduc ion a ou di e en eloci y a es.
The pe o mance o he p oposed p ocedu es has been e alua ed h ough cyclic es s ca -
ied ou wi h di e en eloci y ac o s. The inpu employed o hese es s was he displace-
men ime-his o y measu ed du ing a seismic e en .
The analysis o he esul s allows he ollowing conclusions o be d awn:
•
Remo ing sys ema ic e o s om he expe imen al da a is essen ial o ob ain an accu a e
calib a ion o he co ec ion algo i hms.
•
The i s na u al equency o he co esponding unp o ec ed (wi hou luid iscous
dampe s) s uc u e cons i u es an app op ia e equency o he p oposed ime-his o y
displacemen pa e n o es he dampe s. I s alue can be easily es ima ed in he design
s age h ough nume ical models o ob ained by modal es ing on he ac ual unp o ec ed
s uc u e.
•
I is shown ha o de ze o o he ou pu a iable and o de one o he inpu a iables is
he app op ia e con igu a ion o he il e algo i hm. This solu ion cons i u es a ade-o
be ween accu acy o he p edic ions and ime o p ocessing. Conce ning he algo i hm
based on neu al ne wo ks, he app op ia e numbe o neu ons o he hidden laye is ound
o be ou .
•
The applica ion o he ained algo i hms o seismic expe imen al da a e eals ha bo h
he p oposed algo i hms ou pe o m he simple linea homologa ion in all he analysed
cases. The bes i ing is always achie ed wi h he neu al ne wo k algo i hm. Mo eo e ,
as his algo i hm includes he eloci y ac o as an inpu , i can be used o any es ing e-
loci y wi hin he aining ange, and i e en allows he change o his pa ame e du ing
he es .
REFERENCES
[1]
J. Donea, G. Magone e, P. Neg o, A. Pin o, G. Ve zele i, Pseudodynamic Capabili ies
o he ELSA Labo a o y o Ea hquake Tes ing o La ge S uc u es. Ea hquake Spec-
a (1996) 163-180.
[2]
F. J. Molina, G. Magone e, Pseudodynamic Simula ion o Base Isola ion on a Rein-
o ced Conc e e Building by Means o Subs uc u ing, P oceedings o he Fi s Eu ope-
an Con e ence on S uc u al Con ol, 1996.