Seismic risk assessment of structures by means of stochastic simulation techniques [Elektronische Ressource] / by Enrico Sibilio

Informations

• Publié par: technische_universitat_carolo-wilhelmina_zu_braunschweig
• Publié le: 01 janvier 2007
• Nombre de lectures: 12
• Langue: English
• Poids de l'ouvrage: 9 Mo

Extrait

Seismic Risk Assessment of Structures by
Means of Stochastic Simulation Techniques
Dissertation
submitted to, and approved by,
the Department of Architecture, Civil Engineering and Environmental Sciences
of the Technische Universit¨ at
Carolo-Wilhelmina
zu Braunschweig
and
the Faculty of Engineering
Department of Civil Engineering
of the University of Florence
in candidacy for the degree of a
Doktor-Ingenieur (Dr.-Ing.)/
Dottore di Ricerca in Risk Management on the Built
∗)Environment
by
Enrico Sibilio
from Sezze, Italy
Submitted on 31 March 2006
Oral examination on 20 May 2006
Professoral Advisors Prof. Heinz Antes
Prof. Udo Peil
Prof. Claudio Borri
Prof. Dieter Dinkler
2006
*) Either the German or Italian form of the title may be usedThe dissertation is published in an electronic form by the Braunschweig university
library at the address
http://www.biblio.tu-bs.de/ediss/data/Supervisor
Prof. Dr.-Ing. Marcello Ciampoli University of Rome ”La Sapienza”
Tutors
Prof. Dr.-Ing. Gianni Bartoli University of Florence
Prof. Dr. rer. nat. Heinz Antes Technical University of Braunschweig
Doctoral course coordinators
Prof. Dr.-Ing. Claudio Borri University of Florence
Prof. D Udo Peil Technical University of Braunschweig
Examining Committee
Prof. Dr.-Ing. Marcello Ciampoli University of Rome ”La Sapienza”
Prof. Dr. rer. nat. Heinz Antes Technical University of Braunschweig
Prof. Dr.-Ing. Andrea Vignoli University of Florence
Prof. D J¨ orn Pachl Technical University of Braunschweig
Prof. Dr.-Ing. Renzo Ciuffi University of Florence
Prof. Dr.-Ing. Joachim Stahlmann Technical University of BraunschweigAcknowledgements
First of all, my deep gratitude goes to my supervisor Prof. Marcello Ciampoli,
who encouraged me with his useful suggestions, comments and revisions of this
Thesis. I wish to thank the Italian coordinator of the International PhD Course
Prof. Claudio Borri because he gave me the freedom to decide the best way to
conclude my research. His reprimands have been also appreciated and I must say
that he was always right! I would like to express my appreciation to Prof. Udo
Peil, Prof. Gianni Bartoli and Prof. Heinz Antes.
I wish to say thank to Prof. James Beck because he gave me the possibility to
spend one year at Caltech as Visiting Research Student. It was really exacting,
interesting and fruitful to work with him in a stimulating ambient like Caltech
is. I have really appreciated his lectures on Bayesian Statistics and Stochastic
Simulation, his classes on Applied Mathematics, his comments on my results, and
his defences of the Bayesian approach. His revisions of my draft versions of the
chapters of this Thesis have always been very clear and precise.
It was also a big honor to work with his research group. I wish to thank Matt
Muto for his matlab codes, Alex Taflanidis for his computer for the simulation of
the last minute, for his hints about when to go home to sleep during the nights
spent in my office (Alex, I’ve never listened your hints! thank you anyway!), Judy
Mitrani for her friendship. My thanks also go to Dr. Keith Porter who shared his
office with me during my stay at Caltech, to Carolina Oseguera and to the other
people at Thomas Franklin Laboratory who helped me over the last months.
During my stay at Caltech, I have met many Physicists, Mathematicians,
Engineers, and Biologists. I will keep in my mind a very good trace of the time
spent together with them. They also deserve to be mentioned for their passion
for the research. They have passed their curiosity and spirit on to me. They have
ever been very encouraging and helpful for accomplishing my aims.
I have to thank Marco, Juri, Chiara, Domenico, Erika, Paola, Monica, Mishia,
Stefano, Stefano, Valerio, Ciro, Simona, Edoardo, Diego. This is just a little
piece of the Italian (more or less) scientific community at Caltech and I’m surely
forgetting many people. I am proud of having been a member of this community.
My thoughts go to my friends in Italy who became ”virtual” friends during my
long stay abroad. One person especially deserves my gratitude, Roberto Savini
who is becoming a famous bridge designer, I should say an expert. His friendship
is one of the certainties of my life.
iiiI wish to dedicate this Thesis to my family.
iiiivAbstract
The assessment of the risk represents a fundamental step in the whole process
of the risk management. By definition, the risk due to any natural event may be
quantified only in probabilistic terms. Furthermore, its probabilistic evaluation
involves several uncertain variables and their role must be carefully analyzed.
In this Thesis, issues related to seismic risk assessment have been addressed.
The first part of the Thesis focuses on the definition of the seismic hazard at
a specific site. It is known that the accurate and quantification of
the seismic hazard, defined as the mean annual rate of exceeding of an intensity
measure, depends on the attenuation relation considered, and then its precise for-
mulation deserves particular attention. Attenuation relation is a mathematical
relation between an intensity measure and any pairs of magnitude and epicentral
distance and depends on some model parameters. Moreover, a probabilistic con-
tent is usually associated with the attenuation law which quantifies its inherent
uncertainty.
In order to estimate a mathematical expression and rationally identify its
uncertainty, a fully probabilistic approach has been proposed. The method is
based on Bayesian Model Updating and Robust Predictive Analysis. By using
the Bayes’ theorem the prior probability density function, usually very broad,
reflecting the initial ignorance about the probabilistic contents of the attenuation
law parameters, is updated exploiting the information contained in some data.
The result is a posterior probability density function for the model parameters.
The data considered in this Thesis consist of many actual earthquake records.
The Bayesian updating problem has been solved by using two advanced Markov
Chain Monte Carlo methods. Finally, the robust predictive analysis has been
implemented to account for all the uncertainties involved in the attenuation law
identification.
The second part of the Thesis deals with the assessment of the probability of
failure of various structural models, both linear and nonlinear, subjected to earth-
quakes. First, a probabilistic structural linear model response has been tackled.
The effect of the structural model uncertainty has been quantified through the
identification of the posterior probability density functions of the structural pa-
rameters. More exactly, a Bayesian Model Updating technique in the frequency
domain with unknown non stationary input has been employed; this approach
is able to identify the uncertainty of the stiffness and structural damping at el-
ement level. The seismic risk or the probability of exceeding of a limit state for
a damaged structure has been evaluated for some damage scenarios. The uncer-
vtainty in the definition of the structural capacity has been also addressed. The
probability of failure has been computed following the Subset Simulation method
which is one of the most advanced and efficient Monte Carlo simulation method
used for structural reliability purposes. The aim is to show how the information
obtained through an identification technique can be used in a general reliability
framework.
In the third part of the Thesis, the seismic risk has been evaluated for two
nonlinear structural models. The results of two different techniques have been
compared. The first technique is the well known IM-based approach which is
typically employed in the probabilistic framework of the Performance Based Seis-
mic Design, whereas the second technique is the Subset Simulation. As a first
application, the Subset Simulation has been implemented for calibrating two dif-
ferent approach to the representation of the seismic hazard, namely the classical
Probabilistic Seismic Hazard Analysis (PSHA) and the stochastic ground mo-
tion approach. This phase allows to manage the uncertainties related to the
action. Once the seismic hazard has been defined in a coherent way, the failure
probability may be estimated. Two structural nonlinear models with degrading
stiffness and strength have been considered as examples on which the procedures
have been tested. Both deterministic and uncertain mechanical features for the
models have been implemented. In this way the effect of the structural model
uncertainty has been successfully investigated.
vi