Reliability Seminars

Reliability Seminars were once-a-week gathering of students, visitors and faculty interested in probabilistic methods. Usually, a presentation was made by one of the participants evoking comments and discussion during and after the presentation. Often the presentation was tentative and on a topic of ongoing research, rather than a finished and polished work. As a result, participants became acquainted with each others' current work and the presenter benefited from the discussion of his/her on-going research. These seminars stopped when I left Berkeley to become the president of the American University of Armenia. This page lists the history of these seminars going back to 2005 when I started recording them.

Date: Wednesday, May 28, 2014
Time: 1:30-3PM
Place: PEER Conference Room (325 Davis Hall)
Topic: Nonlinear system identification and application of identifiability and observability tests
Speaker: Professor Andrew Smyth, Columbia University
Abstract: Following a brief overview of several applied structural monitoring activities in New York City, the presentation traces the application of several Bayesian identification techniques such as the Unscented Kalman Filter (UKF) and variants of the Particle Filter for the identification of nonlinear dynamic systems. The use of the UKF for parameter-state estimation is explored with a view to data fusion where strengths of different sensor types are exploited. Applications in civil engineering where the sensing of acceleration and strain is combined with that of displacement, such as with differential GPS, are considered. The efficiency of algorithms such as the UKF permits the parallel competitive consideration of a variety of candidate models for identification in cases where the model class is unknown. Given the powerful nature of the UKF in particular, the question arises of what must be measureable in order for increasingly complex models to be identified. The observability and identifiability tests available from a wide range of sources in the literature are reviewed for nonlinear system dynamics and applications in structural and mechanical systems are presented.

Date: Thursday, May 22, 2014
Time: 1:30-3PM
Place: PEER Conference Room (325 Davis Hall)
Topic: Optimal sensor placement for operational modal analysis based on maximum expected utility
Speaker: Binbin Li
Abstract: Optimal sensor placement essentially is a decision problem under uncertainty. Maximum expected utility theory provides a mathematical foundation for the selection of optimal designs, and the Bayesian approach provides a coherent framework to combine the prior information and uncertainties regarding unknown quantities. In this presentation, we apply the maximum expected utility and a Bayesian linear model for robust sensor placement aiming at operational modal identification. A two-stage decision approach is proposed: the sampling period, sample size, target modes and sensor property are determined in the first stage, then the number and location of sensors are further optimized by maximizing the expected utility in the second stage. Three types of utility functions are investigated, and the elicitation of prior information is discussed. At last, a sensor design example is offered to help understand and validate the proposed approach.

Date: Wednesday, April 30, 2014
Time: 1:30-3PM
Place: PEER Conference Room (325 Davis Hall)
Topic: A new methodology for development of seismic fragility curves
Speaker: Young Joo Lee
Abstract: Accurate seismic fragility curve has been recognized as a key to effective seismic risk assessment of buildings. Existing methodologies for seismic fragility curve development can be classified into four groups: empirical, judgmental, analytical, and hybrid. Among them, analytical fragility curves are most widely used, and they can be categorized again into two subgroups, depending on whether one is using an analytical function or a simulation technique. Although both types have shown decent performances in many seismic fragility problems, they often over-simplify the given problems in reliability analysis or structural analysis due to their own assumptions. In this presentation, a new methodology for the seismic fragility curve development of buildings is proposed. By coupling sophisticated software packages of reliability analysis (FERUM) and structural analysis (ZEUS-NL), the methodology enables us to obtain more accurate seismic fragility curves for less computational cost. The proposed methodology is applied to a numerical example of a 2D frame structure and the results are compared with those by Monte Carlo simulation, demonstrating that the method derives seismic fragility curves accurately and efficiently.

Date: Wednesday, April 16, 2014
Time: 1:30-3PM
Place: PEER Conference Room (325 Davis Hall)
Topic: A new method for determining design points in reliability analysis for a sequence of thresholds
Speaker: Ziqi Wang
Abstract: A crucial step in FORM, SORM and IS is the determination of the so-called design point for a specific limit state function. This point is usually found by optimization algorithms. Since most algorithms generally involve repeated evaluations of the structural response and its gradient, for nonlinear and complicated structural problems this process can be computationally demanding. In this talk I will present a new gradient-based algorithm, namely the "lambda" method, to find the design point for a specific class of reliability problems. As a special case, the proposed method allows solving the reliability problem for a specified reliability index. This property makes the method potentially useful for inverse reliability analysis and performance-based design problems. Preliminary examples demonstrate superior efficiency of the method over the iHLRF algorithm.

Date: Wednesday, April 9, 2014
Time: 1:30-3PM
Place: PEER Conference Room (325 Davis Hall)
Topic: Risk analysis of fatigue-induced sequential failure
Speaker: Young Joo Lee
Abstract: Many structures are subjected to the risk of failures caused by repeated loadings during their service. A challenge in such risk quantification is that a sophisticated structural system with a high level of structural redundancy often has many potential sequences of local failures, which makes it intractable to consider all possible sequences for evaluating the corresponding failure probabilities. In addition, prior local failures that may initiate sequential component failures resulting in a system failure usually re-distribute the loadings in the system, which makes the risk quantification more challenging. In this presentation, the Branch-and-Bound method employing system reliability Bounds (termed the B3 method) which was recently developed for the risk analysis of fatigue-induced sequential failure will be introduced. The B3 method identifies critical sequences of fatigue-induced failure in the decreasing order of their likelihood, as it systematically updates both lower and upper bounds of the system failure probability. After the introduction of the method, it will be demonstrated by numerical examples including an aircraft longeron system.

Date: Wednesday, March 12, 2014
Time: 1:30-3PM
Place: PEER Conference Room (325 Davis Hall)
Topic: Variational Bayesian approach for operational modal identification
Speaker: Binbin Li
Abstract: In operational modal identification, modal parameters obtained are subject to bias and uncertainties. Many identification algorithms have been developed for the most probable value estimation, but only a few attempt to tackle the uncertainty issue. In this presentation, I will show a variational Bayesian approach for operational modal identification, which provides the whole distribution of the modal parameters. Working with the state space representation of a dynamical system, an exact updating formula of the Bayesian state-space model is intractable. Instead, a mean-field approximation of the log evidence is applied, and the Bayesian updating becomes analytically feasible by assuming suitable conjugate priors. The principle and implementation tricks will be introduced. The approach will be illustrated by means of a simulation and a real measured example (operational modal identification of One Rincon Tower).

Date: Wednesday, March 5, 2014
Time: 1:30-3PM
Place: PEER Conference Room (325 Davis Hall)
Topic: Bayesian network methods for modeling and reliability assessment of infrastructure systems
Speaker: Iris Tien
Abstract: The Bayesian network (BN) is an ideal tool for modeling and assessing the reliability of civil infrastructure, particularly when information about the system and its components is uncertain and evolves in time. The major obstacle to the widespread use of BNs for system reliability analysis, however, is the limited size and complexity of the system that can be tractably modeled as a BN. This is due to the exponentially increasing number of elements that must be stored in the conditional probability table (CPT) for the system node in the BN as the number of components in the system increases. In this seminar, I will present novel compression and inference algorithms that I have developed to address this limitation. The algorithms utilize compression techniques to achieve orders of magnitude savings in memory storage for the system CPT. In addition, heuristics developed to improve the computational efficiency of the algorithms are presented. The algorithms are applied to the analysis of an example system, and the performance evaluated compared to existing methods. The gains achieved by the developed algorithms in both memory storage and computation time are demonstrated.

Date: Wednesday, February 19, 2014
Time: 1:30-3PM
Place: PEER Conference Room (325 Davis Hall)
Topic: A probabilistic approach for predicting the resilience of communities
Speaker: James-A. Goulet
Abstract: Given the condition of our infrastructure and the foreseen needs for investments, our communities are known to be vulnerable to large-scale man-made and environmental hazards. The fallback position consists in working on the resilience of communities in order to improve their capacity to recover. Interdependency is a key component of infrastructure resilience. Interdependence may come from spatial proximity, shared resources and also from the dependency of an infrastructure on the services provided by another. This last type of interdependency is the most difficult to handle because it often involves nested complexity. In the case of extreme natural hazards, such as earthquakes or floods, cascading failures are common because of the interdependencies in the services provided by infrastructures. Moreover, the dependency of a community on infrastructure creates an additional challenge where the actors of the recovery are also disrupted. Therefore, our capacity to predict the recovery of cities depends on our capacity to holistically include lifeline infrastructure and societal interdependencies. This seminar will present a probabilistic representation of such social and infrastructure interdependencies. Theoretical examples will highlight the potential of this probabilistic representation for predicting the resilience of communities.

Date: Wednesday, January 29, 2014
Time: 1:30-3PM
Place: 768 Davis Hall
Topic: Extreme value statistics of wind speed data by the ACER method
Speaker: Arvid Naess
Abstract: A novel method for estimating the extreme value statistics associated with a recorded time series of data is presented. In particular, the focus is on recorded wind speed data at some chosen measurement stations along the Norwegian coast. The new method, briefly referred to as the ACER method, is presented in some detail. It is pointed out that the ACER method, if properly implemented, will provide a representation of the exact extreme value distribution given by the data. It is also applicable to nonstationary time series. The estimated return level values obtained by the ACER method are compared to the corresponding values obtained by the annual maxima method and the peaks-over-threshold method. Based on the overall performance, it is concluded that the ACER method may indeed be the preferred choice over the other two methods tested.

Date: Wednesday, December 11, 2013
Time: 1:30-3PM
Place: PEER Conference Room (325 Davis Hall)
Topic: The challenges of high-dimension reliability problems and a new linearization method
Speaker: Marco Broccardo
Abstract: In this talk we will explore the difficulties in solving high-dimensional reliability problems by the first-order reliability method, FORM. We will introduce the concept of the important ring and its role in the computation of failure probability. Since the tail-equivalent linearization method, TELM, is based on FORM and is a high-dimension problem, one must be aware of potential errors when using this method. We present a new type of linearization suitable for high dimension problems, which aims to solve the drawbacks of the original FORM. An application to TELM will be demonstrated.

Date: Wednesday, December 4, 2013
Time: 1:30-3PM
Place: PEER Conference Room (325 Davis Hall)
Topic: Multiple-support response-spectrum (MSRS) method using load-dependent Ritz vectors
Speaker: Ziqi Wang
Abstract: The mode superposition method based upon load-dependent Ritz (LDR) vectors is reviewed and its application on MSRS combination rule is investigated. Two different LDR vector generation methods are explored and their corresponding accuracy for MSRS combination rule are studied. A numerical example of Auburn Ravine bridge will illustrate the method and conclude the presentation.

Date: Wednesday, November 20, 2013
Time: 1:30-3PM
Place: PEER Conference Room (325 Davis Hall)
Topic: Modeling the joint probability of earthquake, site and ground-motion parameters using Bayesian networks
Speaker: Nicolas Kuehn
Abstract: In probabilistic seismic hazard analysis, one wants to calculate the probability of exceedance of a certain destructive ground-motion level at a particular site, which requires a good understanding of the parameters that can influence seismic ground-motions. Bayesian networks are used to investigate the probabilistic dependencies between several parameters of the ground-motion domain. Problems specific to the domain are discussed.

Date: Wednesday, November 13, 2013
Time: 1:30-3PM
Place: PEER Conference Room (325 Davis Hall)
Topic: A data-driven path to post-earthquake rapid structural condition assessment
Speaker: James-A. Goulet
Abstract: Earthquake-prone regions, such as the San Francisco Bay Area, could take advantage of affordable and scalable data-driven strategies for post-earthquake rapid condition assessment. In the literature, several full-scale experiments have provided evidence of a strong correlation between the damage sustained by buildings and the shift in natural frequencies before and after an earthquake. Therefore, monitoring the natural frequencies of a population of sparsely instrumented structures could enable (1) the prioritization of post-earthquake emergency inspections, (2) updating of our knowledge of the state of the entire population as we inspect individual structures, and (3) a rapid classification of structures as "safe" or "not safe." Achieving this requires a change in the way we use data; rather than seeking to infer exact structural condition through complex physics-based models, we need to search for patterns and relationships in the responses of a population of structures. This presentation shows how Bayesian updating can combine the prior knowledge from previous experiments with new information coming from inspections made hours after an earthquake. The expected performance for data-driven post-earthquake rapid structural condition assessment is measured by the percentage of structures assessed 48h after an earthquake and by the percentage of misclassified structures. Probabilistic estimates of these performance metrics are computed for the city of San Francisco using earthquake-induced damage scenarios.

Date: Wednesday, November 6, 2013
Time: 1:30-3PM
Place: PEER Conference Room (325 Davis Hall)
Topic: Adaptive optimization methods in system-level bridge management
Speaker: Haotian Liu
Abstract: We present an adaptive optimization approach, known as Open-Loop Feedback Control (OLFC), for maintenance, repair and replacement planning of systems of bridge components. The proposed implementation of OLFC in Bridge Management Systems is intended to improve bridge management decision-making and deterioration model learning. The OLFC approach is capable of providing more accurate models than the state-of-the-art methods and yielding system cost-savings over any planning horizon when condition survey data are used to update the bridge component deterioration models. OLFC also enables agencies to consider different model classes when learning deterioration models. To illustrate the desirability of this approach, we consider a planning agency managing a system of facilities with limited prior knowledge of the deterioration models over a designated planning horizon. We show that OLFC improves model accuracy and reduces system costs, with a demonstration of how to incorporate system budget constraints when the system is heterogeneous. We focus our discussion on bridge decks, the component of bridge structures that undergoes the fastest deterioration, but the methodology presented in this paper is applicable to all bridge components.

Date: Wednesday, October 30, 2013
Time: 1:30-3PM
Place: PEER Conference Room (325 Davis Hall)
Topic: Shape free stochastic pulse for simulation of ground motions in frequency domain
Speaker: Marco Broccardo
Abstract: Firstly, the representation and the simulation of stochastic processes in frequency domain will be reviewed. Then, preliminary work on the simulation of a shape free random pulse will be presented. This work is particularly suitable for simulations of pulse like synthetic ground motions, a subject extensively studied by Mayssa Dabaghi. In the original work the shape of the pulse is fixed, while in this version this constraint is relaxed. I will explain why I believe the frequency domain representation is particularly suitable to this class of problems. Because of the embryonic state of this work, I would like to receive feedback and sharp critiques in particular from a seismologic prospective.

Date: Wednesday, October 16, 2013
Time: 1:30-3PM
Place: PEER Conference Room (325 Davis Hall)
Topic: Outlier-resistant online structural identification
Speaker: He-Qing (Max) Mu
Abstract: In this presentation, a novel outlier-resistant extended Kalman filter (OR-EKF) is proposed for outlier detection and robust online structural parametric identification using dynamic response data possibly contaminated with outliers. Instead of definite judgment on the outlierness of a data point, the proposed OR-EKF provides the probability of outlier for the measurement at each time step. By excluding the identified outliers, the OR-EKF ensures the stability and reliability of the estimation. In the illustrative examples, the OR-EKF is applied to parametric identification for structural systems with time-varying stiffness in comparison with the plain EKF. The structural response measurements are contaminated with outliers in addition to Gaussian noise. The proposed OR-EKF is capable of outlier detection and it can capture the degrading stiffness trend with more stable and reliable results than the EKF.

Date: Wednesday, October 9, 2013
Time: 1:30-3PM
Place: PEER Conference Room (325 Davis Hall)
Topic: Detectability of anomalies in the context of long-term structural health monitoring - A case study
Speaker: James-A. Goulet
Abstract: This seminar presents the results of a study where real monitoring data recorded on a suspension bridge is used to examine the extent to which anomalies can be detected in the context of long-term structural health monitoring. The study validates that forecasting the performance of a structure using dynamic Bayesian network can detect anomalies having a small magnitude (around 1%), while avoiding false alarms. This validation study has been performed in the context of large-scale civil structures, where missing data and large outliers are common. In addition, an example shows the benefits of using current as well as previous data for forecasting anomalies rather than only using the data available at a current time slice. The second part of the study shows that with the proposed methodology it is possible to detect anomalies having an effect greater than 1% of the structural responses measured, while having a false alarm rate of 0.8 per 10 years.

Date: Wednesday, October 2, 2013
Time: 1:30-3PM
Place: PEER Conference Room (325 Davis Hall)
Topic: Assessing the impact of uncertain structural parameters in inferring maximum structural response using dynamic Bayesian network
Speaker: Iris Tien
Abstract: Previously, I presented a method using dynamic Bayesian network (DBN) to infer the maximum structural response of a building to seismic load. The method processes information from accelerometer recordings on the building and utilizes the Kalman filter to update estimates of the system state. In this talk, I investigate the impact of uncertainty in the structural parameters in performing inference on the structural system. Results of varying the level of uncertainty in the parameters, as well as a method for weighting the inference results by utilizing the likelihood of a sampled parameter are presented.

Date: Wednesday, September 25, 2013
Time: 1:30-3PM
Place: PEER Conference Room (325 Davis Hall)
Topic: Study and selection of ground motion processing techniques for use on simulated ground motions
Speaker: Mayssa Dabaghi
Abstract: In this presentation I review different filtering and baseline correction techniques that have been used to process recorded ground motion acceleration time series. I then illustrate and study their use on simulated ground motions.

Date: Wednesday, September 18, 2013
Time: 1:30-3PM
Place: PEER Conference Room (325 Davis Hall)
Topic: Probabilistic outlier analysis for linear regression and online structural identification
Speaker: He-Qing (Max) Mu
Abstract: Structural identification using dynamic response measurement has attracted tremendous attention over the last decades. In real measurements, outliers exist due to extraordinarily large measurement error and/or modelling error. Since outliers may lead to undesirable identification results, detection and special treatment of outliers becomes an important task in system identification. In this topic, we will introduce a novel concept of outlier probability for outlier detection. The outlier probability is a function of the data size and its normalized residual. The normalized residual is defined as the difference between the measured value and the model output, normalized by its standard deviation. First, we utilize this idea for outlier analysis of linear regression problem. Then, we extend this idea and propose an outlier-resistant extended Kalman filter (OR-EKF) for robust structural parameter identification using dynamic response data. In contrast to other existing outlier detection criteria that require some subjective bound (e.g., normalized residual larger than 2.5), the outlier probability threshold of 0.5 is intuitive. Finally, applications to online structural identification using outlier-contaminated dynamic measurements will be presented.

Date: Wednesday, September 11, 2013
Time: 1:30-3PM
Place: PEER Conference Room (325 Davis Hall)
Topic: Optimal Planning and Learning for Management of Wind Farms
Speaker: Matteo Pozzi
Abstract: Wind energy is a key renewable energy source, yet wind farms have relatively high cost compared to many traditional energy sources. Among the life cycle cost of wind farms, operation and maintenance (O&M) accounts for 25-35%, and an efficient strategy for management of turbines can significantly reduce the O&M cost. Traditional methods to model the O&M process, as Markov Decision Processes (MDPs) and Partially Observable MDPs (POMDPs), have limitations that do not allow the model to include properly the knowledge available, and which may result in non-optimal strategies for management of the farm. Specifically, the conditional probabilities for modeling the degradation process and the precision of the observations are usually affected by epistemic uncertainty. While MPDs and POMDPs are formulated for fixed transition and observation probabilities, the Bayes-Adaptive POMDP (BA-POMDP) framework treats those conditional probabilities as random variables, and is therefore suitable for including epistemic uncertainty. In this seminar, I present a novel learning and planning method, called Planning and Learning in Uncertain dynamic Systems (PLUS), within the BA-POMDP framework, that can learn from the environment, update the distributions of model parameters, and select the optimal strategy considering the uncertainty related to the model.

Date: Wednesday, September 4, 2013
Time: 1:30-3PM
Place: PEER Conference Room (325 Davis Hall)
Topic: Pre-posterior optimization of sequence of measurement and intervention actions under structural reliability constraint
Speaker: James-A. Goulet
Abstract: It is common to assess the condition of an existing infrastructure using reliability analysis. When, based on the available information, an existing structure has an estimated failure probability above the admissible level, the default solution often is to either strengthen or replace it. Even if this practice is safe, it may not be the most economical. In order to economically restore and improve our existing infrastructure, the engineering community needs to be able to assess the potential gains associated with reducing epistemic uncertainties using measurements, before opting for costly intervention actions, if they become necessary. A pre-posterior analysis framework is proposed to (1) optimize sequences of actions minimizing the expected costs and satisfying reliability constraints, and (2) quantify the potential gain of making measurements in existing structures. Illustrative examples show that when the failure probability estimated based on the present state of knowledge does not satisfy an admissible threshold, strengthening or replacement interventions can be sub-optimal first actions. The examples also show that significant savings can be achieved by reducing epistemic uncertainties.

Date: Wednesday, August 28, 2013
Time: 1:30-3PM
Place: PEER Conference Room (325 Davis Hall)
Topic: Response-spectrum compatible power-spectral density for high-frequency range
Speaker: Armen Der Kiureghian
Abstract: A response-spectrum-compatible power spectral density (PSD) is used in design of nuclear power plants to circumvent costly time-history analysis. Several methods are available for identifying the compatible PSD. However, investigators have recently noted that compatibility between the response spectrum and the corresponding PSD may be lacking in the high frequency range. I will examine three methods for generating a PSD corresponding to a response spectrum that is compatible for the entire range of frequencies. I will also discuss conditions on the response spectrum shape to assure this level of compatibility.

Date: Thursday August 8, 2013
Time: 3-4:30PM
Place: 786 Davis Hall
Topic: Service life prediction of reinforced concrete structures exposed to corrosion
Speaker: Imad El Khouri
Abstract: Application of Fick's Second Law of diffusion to simulate the penetration of chlorides into the concrete. The Lightcon model -a deterministic tool- developed in the 1990s uses Fick's law to give a specific time to corrosion initiation. However the concrete and the environment are very random and thus require a stochastic approach. Four parameters are assigned distributions (taken from literature) then using monte carlo and modified monte carlo method for sampling, we produce a distribution for the time of corrosion initiation. In addition to that we analyse the factor with the highest effect on the result. Comparison with other deterministic life predicting tools is yet to be done using the same procedure.

Date: Thursday, August 1, 2013
Time: 3-4:30PM
Place: PEER Conference Room (325 Davis Hall)
Topic: Dynamic Bayesian network for operational modal identification
Speaker: Binbin Li
Abstract: Application of the dynamic Bayesian network (DBN) for modal identification under operational conditions is investigated. Under the assumption of Gaussian white noise, the stochastic state space model is converted to a DBN. Time-domain maximum likelihood estimation (MLE) is applied to learn the DBN. The expectation-maximization (EM) algorithm, initialized by parameters identified by the reference-based, data-driven stochastic subspace identification (SSI/DATA) method, is used to iteratively solve for the MLE. A numerical simulation example demonstrates the performance of the DBN-based identification method.

Date: Thursday, July 25, 2013
Time: 3-4:30PM
Place: PEER Conference Room (325 Davis Hall)
Topic: Forecasting anomalies using monitoring data and dynamic Bayesian networks
Speaker: James-A. Goulet
Abstract: Structural health monitoring (SHM) systems often rely on physics-based models for interpreting data and sending alerts when preventive interventions are required. In the context of large-scale civil structures, building unbiased physics-based models is a resource-demanding challenge. Alternative signal processing techniques have been developed for identifying anomalies using SHM data without relying on complex physics-based models. Aspects limiting the widespread use of these interpretation methods are related to (1) extending the scope of SHM to anomaly detection, (2) handling missing data and outliers, (3) identifying trade-offs between anomaly detection capacity and false-alarm rate. This presentation describes a data-driven probabilistic framework able to address the current limitations (1-3). Here, anomalies do not necessarily refer to discrete damage events but more generally to any gradual deterioration or malfunction either in the structure or in the monitoring system. The methodology is formulated as a dynamic Bayesian network providing a real-time assessment as well as forecasts of the probability that a structure is and will be in an abnormal state.

Date: Wednesday, June 12, 2013
Time: 3:30-5PM
Place: PEER Conference Room (325 Davis Hall)
Topic: Dry runs for ICOSSAR 2013 presentations:

Mayssa Dabaghi: "Seismic hazard analysis using simulated ground motions"
Abstract: In the current engineering practice, seismic hazard is characterized by a target response spectrum obtained from a probabilistic seismic hazard analysis (PSHA) at the site of interest, using ground motion prediction equations (GMPEs). The spectrum is then used to compute structural responses. This approach is only adequate for elastic structures located far from the fault rupture. In this paper, we illustrate the use of simulated ground motions instead of GMPEs to perform PSHA, where the ground motions are simulated using parameterized stochastic ground motion models developed by Rezaeian and Der Kiureghian (2010) and Dabaghi et al. (2011). Response history analysis is used and the hazard is calculated for a simple structure idealized as an inelastic, single-degree-of-freedom oscillator.

Marco Broccardo: "Non-stationary stochastic dynamic analysis by tail-equivalent linearization"
Abstract: The tail-equivalent linearization method (TELM) is a recently developed computational method to solve nonlinear stochastic dynamic problems by the first-order reliability method (FORM). TELM employs a tail-equivalent linear system (TELS) by equating the tail probability of a linear system to the first-order approximation of the tail probability of the nonlinear system. For stationary problems, the TELS is time-independent and only one linear system needs to be defined to study the statistics of the response. However, for a transient input, the TELS is time-dependent. Thus, TELSs for different time points must be defined to study the non-stationary response. Since each TELS is obtained from the solution of an optimization problem, the computational cost required to solve the non-stationary problem can be prohibitive. This paper tackles the class of non-stationary problems described via evolutionary power spectral density by defining an evolutionary TELS (ETELS) in place of a series of point-in-time TELSs. An example shows the accuracy and effectiveness of the method.

Iris Tien: "Compression algorithm for Bayesian network modeling of binary systems"
Abstract: A Bayesian network (BN) is a useful tool for analyzing the reliability of systems. The BN framework is limited, however, by the size and complexity of the system that can be tractably modeled. Each node in a BN graph is associated with a conditional probability table (CPT), the size of which grows exponentially with the number of connected nodes in the graph, presenting a memory storage challenge in constructing and analyzing the BN. In this paper, we look at binary systems, where components of the system are in either one of two states, survival or failure, and the component states deterministically define the system state. This analysis is particularly useful for studying the reliability of infrastructure systems, where, e.g., the states of individual gas pipelines or roads directly impact the state of the overall natural gas or transportation system. We present a compression algorithm for the CPTs of such systems so that they may be modeled on a larger scale as BNs. We apply our algorithm to an example system and evaluate its performance compared to an existing algorithm.

Date: Thursday, June 6, 2013
Time: 1:30-3PM
Place: PEER Conference Room (325 Davis Hall)
Topic: Dry runs for ICOSSAR 2013 presentations:

Iris Tien: "Inference on maximum structural response based on measured accelerations using dynamic Bayesian network"
Abstract: The dynamic Bayesian network (DBN) is a useful tool for analyzing uncertain systems that evolve with time. As such, it is useful in structural health monitoring applications where measurements can be noisy and uncertain. We propose an algorithm that uses a DBN formulation of the system to assess the maximum response of a structure under seismic load, given recordings from accelerometers placed on the structure. By investigating the impact on the accuracy of the sensor characteristics, including the number of sensors, their positions within the structure, and the precision of their measurements, the results of our analysis inform decision making on the use of accelerometers in the monitoring of structures under seismic loads. We assume linear Gaussian response, which is appropriate for serviceability studies. An example application demonstrates that the DBN is a powerful tool for processing of information gained from monitoring devices.

Binbin Li: "Optimal sensor placement for structural health monitoring based on K-L divergence"
Abstract: The K-L divergence criterion in optimal sensor placement (OSP) for damage identification is investigated. The K-L divergence criterion aims at minimizing the distance between the probability distributions of identified damage by using all potential measuring points and that by utilizing partial ones. An asymptotic approximation of the probability distribution of damage parameters and expectation of K-L divergence are applied successively to make it feasible in implementation. It is shown that the K-L divergence is a versatile criterion which corresponds to the criteria of Fisher information, information entropy and Mahalanobis distance in its different types. In addition, an improved version of cross-entropy (CE) method is introduced to solve this binary combinatorial optimization problem. Finally, OSP problems of shear-type building and Aizhai suspension bridge are used to illustrate the validity of the K-L divergence criterion and the CE method.

James-A. Goulet: "Probabilistic model falsification for structural identification"
Abstract: Model-based structural identification uses physics-based models to infer the property of struc- tures. Several types of structural identification approaches are used in the context of large-scale infrastructure diagnosis. These approaches are often either highly subjective or require the fulfillment of restrictive hypothe- ses related to the definition of uncertainties. This paper presents a probabilistic model falsification methodology for structural identification. Instead of looking for model parameters that best match measurement data, this methodology falsifies models that are not compatible with measurements. The methodology is most appropriate for large-scale infrastructure applications where little information is available to define modeling uncertainties.

Date: Thursday, May 30, 2013
Time: 1:30-3PM
Place: PEER Conference Room (325 Davis Hall)
Topic: Stochastic analysis of precipitates in a material grain
Speaker: Armen Der Kiureghian
Abstract: Ni-based alloys have dual phase microstructure consisting of a face-centered-cubic (FCC) matrix and Ni3Al precipitates with Ll2 structure embedded coherently within the matrix. The yield strength of the alloy depends on the number and shape of the intersections of precipitates with the {1,1,1} plane of the FCC crystals. In this paper, we develop closed form solutions for the statistics of the precipitate intersections with the {1,1,1} plane of the FCC for precipitates of cubic and spherical shapes. Results are derived for the number of intersections, shapes of intersection areas, and moments and distributions of intersection areas. Results for both uniform and variable size precipitates are given. These results can be used in predicting the yield strength of the alloy.

Date: Thursday, May 23, 2013
Time: 1:30-3PM
Place: PEER Conference Room (325 Davis Hall)
Topic: Progress in the studies of non-stationary stochastic dynamic analysis by tail-equivalent linearization
Speaker: Marco Broccardo
Abstract: Last November I presented the theory underlying the solution of nonstationary dynamics problem by use of TELM analysis. This talk presents the evolution of that original work that will be officially presented at ICOSSAR 2013. The original Tail-Equivalent Linearization Method (TELM) is a computational method that solves nonlinear stochastic dynamic problems by the First-Order Reliability Method (FORM). For stationary problems TELM is well established, however for a transient input, the original method requires multiple TELM analyses. TELM analysis demands the solution of a nonlinear optimization problem; therefore, for non-stationary problems, multiple solutions can become unfeasible or too costly. This talk aims to solve the class of non-stationary problems described by Priestley's evolutionary power spectral density, by defining an evolutionary TELM analysis in place of a series of point-in-time TELM analyses. Only one solution of the optimization problem is required to define the evolutionary analysis.

Date: May 1, 2013
Time: 1:30-3PM
Place: PEER Conference Room (325 Davis Hall)
Topic: Uncertainty quantification using exponential epi-splines
Speaker: Johannes Royset
Abstract: We quantify uncertainty in complex systems by a flexible, nonparametric framework for estimating probability density functions of output quantities of interest. The framework systematically incorporates soft information about the system from engineering judgment and experience to improve the estimates and ensure that they are consistent with prior knowledge. The framework is based on a maximum likelihood criterion, with epi-splines facilitating rapid solution of the resulting optimization problems. In four numerical examples with few realizations of the system output, we identify the main features of output densities even for nonsmooth and discontinuous system function and high-dimensional inputs.

Date: April 24, 2013
Time: 1:30-3PM
Place: PEER Conference Room (325 Davis Hall)
Topic: Life-Cycle Maintenance Management for Bridge Networks under Reliability Consideration
Speaker: Xiaofei Hu
Abstract: The research addresses the problem of optimizing maintenance, rehabilitation and replacement (MR&R) decisions for deteriorating bridges in highway networks. The objective is to ensure an adequate level of network reliability at the lowest possible life-cycle maintenance cost. An analytical and efficient approach to the network-level problem is provided. We show that the expression for network reliability level can be obtained directly from the minimal cut sets under reasonable approximations. Instead of considering the evolution of network reliability over the whole lifecycle, the optimization is formulated with a constraint on the lower bound of network reliability, which reduces the network-level problem to one of optimizing the set of reliability levels for individual bridges. For simple decomposable networks, the network-level optimization can be tackled with standard techniques in polynomial time. For complex networks which are indecomposable, we provide an algorithm to transform the nonlinear integer programming problem into binary integer linear programming. Finally, the framework is extended to networks with multiple origin-destination pairs.

Date: April 17, 2013
Time: 1:30-3PM
Place: PEER Conference Room (325 Davis Hall)
Topic: Incorporating Network Considerations into Pavement Management Systems
Speaker: Aditya Medury
Abstract: The objective of infrastructure management is to provide optimal maintenance, rehabilitation and replacement (MR&R) policies for a system of facilities over a planning horizon. While most approaches in the literature have studied the decision-making process as a finite resource allocation problem, the impact of construction activities on the road network is often not accounted for. The state-of-the-art Markov decision process (MDP)-based optimization approaches in infrastructure management, while optimal for solving budget allocation problems, become internally inconsistent upon introducing network constraints. In comparison, approximate dynamic programming (ADP) enables solving complex problem formulations by using simulation techniques and lower dimension value function approximations. In this paper, an ADP framework is proposed, wherein capacity losses due to construction activities are subjected to an agency-defined network capacity threshold. A parametric study is conducted on a stylized network configuration to infer the impact of network-based constraints on the decision-making process.

Date: April 10, 2013
Time: 1:30-3PM
Place: PEER Conference Room (325 Davis Hall)
Topic: Seismic hazard analysis using simulated ground motions
Speaker: Mayssa Dabaghi
Abstract: In the current engineering practice, seismic hazard is characterized by a target response spectrum obtained from a probabilistic seismic hazard analysis (PSHA) at the site of interest, using ground motion prediction equations (GMPEs). The spectrum is then used to compute structural responses. This approach is only adequate for elastic structures located far from the fault rupture. In this paper, we illustrate the use of simulated ground motions instead of GMPEs to perform PSHA, where the ground motions are simulated using parametrized stochastic ground motion models developed by Rezaeian and Der Kiureghian (2010) and Dabaghi et al. (2011). Response history analysis is used and the hazard is calculated for a simple structure idealized as an inelastic, single-degree-of-freedom oscillator.

Date: Wednesday April 3, 2013
Time: 1:30-3PM
Place: 786 Davis Hall
Topic: Value-of-information analysis using structural reliability methods
Speaker: Daniel Straub
Abstract: As is well known, the value-of-information (VoI) concept allows one to quantify the benefit of an inspection or monitoring activity or any other means of obtaining information and reducing uncertainty This is of relevance e.g. when designing and structural health monitoring system or when planning inspections of structures. In this talk, I will present the VoI concept from the point-of-view of structural reliability methods and explore the advantages of such an interpretation. I will show some initial application examples to demonstrate the concepts. In the second part of the talk, I will discuss the extension of the concept to more complex real decision situations, illustrated through several examples of inspection planning in offshore and marine structures and monitoring of aircraft structures.

Date: March 22, 2013
Time: 11-12:30PM
Place: 786 Davis Hall
Topic: Anomaly detection using Bayesian data analysis - Advancement report
Speaker: James-A. Goulet
Abstract: This presentation will be an advancement report regarding the development of methods for assessing the state of structures or agents using Bayesian data analysis. For a first application on construction-site hazard mitigation, this presentation will show how this problem can be formulated using dynamic Bayesian networks and how Kalman filters can be used to update the state of the system. Parameters involved in the anomaly detection are calibrated in order to maximize the expected utility of the system. In this case, maximizing the expected utility involves minimizing the occurrence of false alarms and the detection time. For a second example involving structural health monitoring, this presentation will show how the problem can be formulated as a dynamic Bayesian network. This formulation will be compared with existing data analysis methodologies.

Date: March 15, 2013
Time: 11-12:30PM
Place: PEER Conference Room (325 Davis Hall)
Topic: Assessing maximum structural response under seismic load using dynamic Bayesian network
Speaker: Iris Tien
Abstract: A Dynamic Bayesian network (DBN) is a statistical tool used to perform prediction and inference in systems evolving with time. The Kalman smoother (KS) is a classical framework for the processing of dynamic data in linear Gaussian models, which can be represented as DBNs. In this presentation, the KS is applied to model the response of a structure under seismic excitation. The goal is to infer the maximum response during the earthquake, considering prior knowledge on the ground motion and recordings of accelerometers mounted on the structure. An application of the proposed algorithm will be presented using a shear-type building model, with the goal of estimating the maximum interstory drift. I will investigate the impact on the accuracy of the estimation of varying the characteristics of the measurements used in the analysis, including the number of sensors, their positions within the structure, and the precision of their measurements. In addition, both analytical formulas and Monte Carlo simulations will be utilized to obtain and compare the distributions of the maximum interstory drift under seismic load.

Date: March 8, 2013
Time: 11-12:30PM
Place: 786 Davis Hall
Topic: Earthquake scenarios at a Berkeley site
Speaker: Mayssa Dabaghi
Abstract: In this presentation, I will present the procedure for building a representative set of earthquake scenarios for a site located in Berkeley, only accounting for the Hayward.

Date: March 1, 2013
Time: 11-12:30PM
Place: PEER Conference Room (325 Davis Hall)
Topic: Dynamic Bayesian network for system identification
Speaker: Binbin Li
Abstract: Dynamic Bayesian networks (DBNs) generalize state-space models by representing hidden and observed states as state variables. The graphical structure of DBNs provides an easy way to specify, interpret and estimate these conditional interdependencies. After converting the differential equation governing the structural vibration into a DBN, it becomes straightforward to apply the learning algorithm to identify modal parameters. A framework based on DBN is proposed for system identification, which is applicable to both time-invariant and time-variant systems. For illustration, we focus on the modal identification of linear dynamic structures with output-only data. The modal parameters encoded in the state transition matrix are identified by maximum likelihood estimation under the assumption of Gaussian noise. The effectiveness of the proposed DBN is investigated by a numerical example of a building structure with multiple degrees of freedom.

Date: Friday 22, 2013
Time: 11-12:30PM
Place: 615 Davis
Topic: Multi-Component Nonlinear Stochastic Dynamic Analysis Using TELM
Speaker: Marco Broccardo
Abstract: In this seminar I present the extension of the Tail-Equivalent Linearization Method, TELM, to the case of a nonlinear structure subjected to multiple stochastic excitations. After a brief review of the basis of TELM, the extension to the multi-component case is explained through its application to an asymmetric one-story building with hysteretic behaviour. A range of eccentricities controls the asymmetry and the effect on the probability of failure of the system is studied in detail. The results obtained by TELM are in close agreement with Monte Carlo simulation results.

Date: Wednesday January 30, 2013
Time: 1:30-3PM
Place: 786 Davis
Topic: CQC modal combination rule for floor acceleration
Speaker: Armen Der Kiureghian
Abstract: The peak floor acceleration (PFA) is a critical parameter influencing the performance of non-structural elements in buildings. A CQC-type modal combination rule is developed to compute the PFA, while approximately accounting for the "rigid" contribution of truncated modes. Both discrete and continuous systems are considered. Application of the method to a continuous cantilever beam idealizing a building with shear walls is presented and compared with alternative formulations. The proposed method is able to provide a consistent estimation of the PFA along the entire structure, not only where the PFA is principally influenced by the first few flexible modes, but also where the PFA is mainly related to the rigid response of the structure, for example near the base.

Date: Wednesday January 23, 2013
Time: 2-3:30PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Anomaly detection using sequential Bayesian data-analysis
Speaker: James-A. Goulet
Abstract: When dealing with a large population of structures or agents, the construction of proper behavior models can be cost-prohibitive. Therefore, when it is time to assess the state of these structures or agents, it is more convenient to rely on measurement data only. One of the main challenges associated with this task is to minimize the occurrence of false positive (i.e. false alarms). In this work sequential Bayesian data-analysis is used to detect anomalies and mitigate hazard using behavior measurements. Applications are presented for two contexts: construction-site hazard mitigation, and anomaly detection for structural health monitoring.

Date: Wednesday January 16, 2013
Time: 2-3:30PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Post-earthquake Damages Identification using Bayesian network? A Naiive Example
Speaker: Binbin Li
Abstract: A Bayesian network methodology is developed for real-time identification of post-earthquake damages. This method employs the equivalent earthquake force and identifies modal parameters to get the posterior distribution of all potential damage scenarios -- combinations of plastic hinges. In order to identify damages caused by the main shock and aftershock, static and dynamic Bayesian network models are constructed individually. Monte Carlo sampling and discretization are employed to make inference. In the numerical example, the proposed models successfully detect the potential damage.

Date: Wednesday December 5, 2012
Time: 2-3:30PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Measures of risk, buffered reliability, and their applications to design optimization
Speaker: Johannes O. Royset
Abstract: In reliability engineering focused on the design and optimization of structures, the typical measure of reliability is the probability of failure of the structure or its individual components relative to specific limit states. However, the failure probability has troublesome properties that raise several theoretical, practical, and computational issues. This presentation explains the seriousness of these issues in the context of risk assessment and design optimization, and discusses alternative measures of risk. In particular, we discuss the buffered failure probability, which offers significant advantages. The buffered failure probability is handled with relative ease in design optimization problems, accounts for the degree of violation of a performance threshold, and is more conservative than the failure probability. We also make connections with statistical estimation, which inevitably is an important part of risk analysis of hazards that are only known through a data base of historical events. These connections result in alternative regression techniques that go beyond classical least-squares and that are more suitable for a risk-averse decision maker.

Date: Wednesday November 28, 2012
Time: 2-3:30PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Bayesian inference for identification of structural properties: Challenges associated with using idealized models
Speaker: James-A. Goulet
Abstract: Transportation infrastructure, water distribution networks and energy production systems are now aging displaying serviceability and security issues. Monitoring the responses of structures can help understanding their behavior and quantifying the need for rehabilitation, thereby supporting interventions prioritization if not avoidance. Probabilistic techniques are traditionally used to identify the properties of systems by comparing observations with predictions obtained from physics-based models. There are challenges associated with the identification of properties for large-scale infrastructure. One of them is our reliance on idealized models where the relationships between prediction errors are usually unknown. This seminar presents how, in the context of Bayesian inference, overlooking this aspect can lead to biased diagnoses. It will be shown how likelihood functions can be modified to obtain correct diagnoses in situations where there are aleatory and epistemic uncertainties; this without requiring to know the relationships between prediction errors. These aspects are introduced using an illustrative example and are then put in the context of large-scale infrastructure diagnosis.

Date: Wednesday November 21, 2012
Time: 2-3:30PM
Place: 768 David Hall
Topic: Non-stationary stochastic dynamic analysis by tail-equivalent linearization
Speaker: Marco Broccardo
Abstract: The Tail-Equivalent Linearization Method (TELM) is a computational method developed to solve nonlinear stochastic dynamic problems by the First-Order Reliability Method (FORM). For stationary problems, TELM is well established and depends mainly on a selected threshold for the response of interest and is independent from the time variable once the stationary condition is achieved. However, for a transient input, the analysis also depends on the selected point in time and repeated analyses for a given threshold have to be performed to study the non-stationary response. TELM analysis requires the solution of a nonlinear optimization problem; therefore, for non-stationary problems, multiple solutions can become unfeasible or too costly. This talk aims to tackle the class of non-stationary problems described by Priestley's evolutionary power spectral density, by defining an evolutionary TELM analysis in place of a series of point-in-time TELM analyses. Only one solution of the optimization problem is required to define the evolutionary analysis.

Date: Wednesday November 14, 2012
Time: 2-3:30PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Stochastic simulation of near-fault ground motions: A study of the variability of simulated residual motions
Speaker: Mayssa Dabaghi
Abstract: In this presentation I examine the variability of simulated residual motions having the same residual model parameters. This variability is compared to the variability of the residuals of recorded near-fault ground motions. The residual model used is the non-stationary modulated and filtered white noise model with time-varying filter parameters (Rezaeian and Der Kiureghian, 2010).

Date: Monday November 5, 2012
Time: 2-3:30PM
Place: 768 Davis Hall
Topic: Modeling of infrastructure system reliability using Bayesian networks
Speaker: Iris Tien
Abstract: A Bayesian network (BN) is a probabilistic framework well suited for analyzing the reliability of systems. The BN framework is limited, however, by the size and complexity of the system that can be tractably modeled. Each node in a BN graph is associated with a conditional probability table (CPT), the size of which grows exponentially with the number of connected nodes in the graph, presenting a memory storage challenge in constructing and analyzing the BN. The focus in this presentation will be on binary systems, where components of the system are in one of two states: survival or failure. This analysis is particularly useful for studying the reliability of infrastructure systems, where, e.g., the component states of individual gas pipelines or roads directly impact the state of the overall natural gas or transportation system. A compression algorithm for the CPTs of these systems will be presented, which enables systems to be modeled on a larger scale as BNs. The algorithm is applied to an example system and its performance evaluated compared to existing algorithms.

Date: Friday August 24, 2012
Time: 1:30-3PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Information theoretic optimal sensor placement in structural health monitoring
Speaker: Binbin Li
Abstract: Since most civil structures are large and complicated, it is both infeasible and unnecessary to install too many sensors in order to diagnose deterioration of the structure. Hence, optimal sensor placement has become an important element in design of structural health monitoring systems. In this presentation, an overall, systematic application of this problem is illustrated. First, a thorough description of the Fisher information criterion for sensor placement is presented, including the popular D-optimality and T-optimality criteria. Then, the conditional entropy, which is proved to be the lower bound of error entropy, is applied in the sensor placement aiming at damage identification. Lastly, the K-L divergence criterion, a load and damage dependent criterion, is pioneered in sensor placement. The talk will also briefly introduce several large bridges in China.

Date: Thursday August 16, 2012
Time: 1:30-3PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: The Tail Equivalent Linearization Method in random vibration analysis
Speaker: Marco Broccardo
Abstract: In preparation for my qualifying examination, in this talk I present an overview of the tail-equivalent linearization method (TELM) to solve non-linear random vibration problems. The talk first illustrates the main concepts underling the computational method and the describes recent developments in the subject.

Date: Thursday August 9, 2012
Time: 1:30-3PM
Place: 768 Davis Hall
Topic: Stochastic simulation of near-fault ground motions
Speaker: Mayssa Dabaghi
Abstract: Near-fault ground motions often possess distinct characteristics, which can have strong influence on structural response. Due to scarcity of recorded near-fault ground motions, there is interest in developing synthetic ground motions which can be used in performance-based earthquake engineering in addition to or in place of recorded motions. It is crucial that such synthetic motions be realistic and have characteristics that are consistent with recorded near-fault ground motions. Furthermore, it is most useful if the synthetic motions are generated in terms of information that is normally available to the design engineer, such as the earthquake magnitude, the position of the site relative to the potential fault rupture, and the shear-wave velocity of the soil at the site. I will briefly present the parametrized stochastic model of near-fault ground motion in the strike-normal direction that we developed and I will illustrate its use by generating synthetic near-fault ground motions for a given set of earthquake source and site characteristics.

Date: Thursday August 2, 2012
Time: 1:30-3PM
Place: 768 Davis Hall
Topic: Simulation of stochastic processes in time domain by sinc basis functions and application in TELM analysis
Speaker: Marco Broccardo
Abstract: This talk presents an alternative and comprehensive representation of band-limited stochastic processes. The approach employs a unified framework for such representations by use of concepts of Hilbert space and general decomposition of a signal into basis functions. The proposed representation, in general, allows reducing the number of random variables used in the simulation of stochastic processes. Direct application of the representation is employed to the Tail-Equivalent Linearization Method (TELM) analysis. An example of a single degree of freedom hysteretic oscillator subjected to a Gaussian band-limited white noise simulated by use of sinc basis functions is presented. The accuracy and the efficiency of the new representation are compared with the current discretizations.

Date: Thursday July 26, 2012
Time: 1:30-3PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: The choice of one-dimensional interpolation points in the sparse-grid collocation method
Speaker: Jun He
Abstract: The presentation discusses how to choose one-dimensional interpolation points such that we can obtain complete and incomplete polynomials in the sparse-grid collocation method. The calculation accuracy and computational costs of complete and incomplete polynomials are investigated through a simple numerical example.

Date: Monday May 21, 2012
Time: 3:30-5PM
Place: 542 Davis Hall
Topic: Wavelet-based damage sensitive features for seismic damage diagnosis of civil structures
Speaker: Hae Young Noh
Abstract: After an earthquake, the immediate diagnosis of structural damage is of great importance for facilitating emergency response and preventing further losses and injuries. Similarly, decisions on repair and rehabilitation can be greatly improved by obtaining accurate information on damage. Recent developments in sensor technology and wireless communication systems along with advances in damage diagnosis algorithms have brought us closer to the process of structural damage assessment without conducting visual inspections. In this presentation, I will introduce three wavelet-based damage sensitive features that utilize the acceleration response of a structure recorded during an earthquake. Since earthquake excitations are non-stationary, the wavelet transform, which represents data as a weighted sum of time-localized waves, is used to model the structural responses. The validations using experimental and simulated data show that the values of these features migrate as the damage extent increases. Thus, structural damage can be inferred in a robust manner under various input ground motions.

Date: Thursday May 17, 2012
Time: 3:30-5PM
Place: 542 Davis Hall
Topic: Hazard estimation of existing buildings affected from Kocaeli and Northdridge earthquakes due to local soil conditions
Speaker: Elmon Toraman
Abstract: The Artificial Neural Network (ANN) methodology is used to model complex engineering problems, where the relationship between the input and the output is non-linear. After introducing the ANN methodology, this talk will focus on the soil properties affecting the building damage during earthquakes. The ANN approach is used to identify the cause of the damage based on soil properties excluding the structural effects. Two different earthquakes with similar properties, Kocaeli (1999) and Northridge (1994), are used in the presented models.

Date: Thursday May 3, 2012
Time: 3:30-5PM
Place: 542 Davis Hall
Topic: Consolidating real records selection for earthquake resistant structural design and assessment
Speaker: Carmine Galasso
Abstract: As nonlinear dynamic analysis (NLDA) becomes a more frequently used procedure in research, international codes and practice for assessing the demand on a structure due to earthquakes, ground motion selection becomes increasingly relevant. In order to select representative ground motions to effectively assess structural performance, it is important, first of all, to understand which properties of a recorded ground motion are most strongly related to the response caused in the structure and, consequently, have to take into account. Then, it is important to develop and introduce into the practice of structural engineering, tools which may effectively help to address the basic issues related to seismic input selection, trying to reconcile research achievements and code-based practice. Based on these remarks, the aim of this seminar is to shed some light on a number of aspects involved when selecting and, eventually, manipulating ground motions for NLDA. Practical guidance, conditional hazard maps for secondary intensity measures and software tools are developed and provided to the engineering community for choosing appropriate ground motions in predicting structural response by NLDA.

Date: Thursday April 19, 2012
Time: 3:30-5PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Simulation of near-fault ground motions for specified earthquake source and site characteristics: Predictive equations and simulation examples
Speaker: Mayssa Dabaghi
Abstract: In this presentation, I will first examine the predictive equations for the 12 model parameters, developed using regression analysis. I will then present some simulated near-fault ground motions for various values of earthquake magnitude, source to site distance and site soil conditions. The simulated motions will be compared to the recorded motions in the database.

Date: Thursday April 5 and 12, 2012
Time: 3:30-5PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Bayesian Network with Gaussian variables for post-earthquake emergency management
Speaker: Matteo Pozzi
Abstract: This seminar proposes a probabilistic methodology to assess the reliability of a bridge network in the aftermath of an earthquake, allowing real-time updating when data collected by visual inspection or instrumented monitoring are available. The approach makes use of a Bayesian network with conditional Gaussian distributions to model the correlations in the demands and in capacities of the structures. Its main benefit relative to the use of a Bayesian network with discrete variables is that the approach can handle a large number of structures (in the order of few thousands), performing exact inference. We present the effectiveness of the methodology on a large bridge network, showing how the reliability of the connectivity between selected locations is progressively updated by the information collected in the field about the condition state of selected bridge and the acceleration experienced at recording sites. Finally, we discuss the capability of the system to identify the safest path to any target location, which is of key relevance for the management of the rescue teams.

Date: Thursday March 22, 2012
Time: 3:30-5PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: An Algorithm for Modeling Large Systems as Bayesian Networks
Speaker: Iris Tien
Abstract: A Bayesian network (BN) is a useful tool for analyzing the reliability of systems and for making engineering decisions under uncertainty. The BN framework is limited, however, by the size and complexity of the system that can be tractably modeled. Each node in a BN graph is associated with a conditional probability table (CPT), the size of which grows exponentially with the number of connected nodes in the graph, presenting a memory storage challenge in constructing and analyzing the BN. The algorithm presented in this seminar is based on a minimum cut set/minimum link set (MCS/MLS) formulation of a system, and takes advantage of the special nature of the MCS/MLS CPTs to be able to model large systems as BNs.

Date: Thursday March 15, 2012
Time: 3:30-5PM
Place: 542 Davis Hall
Topic: A stochastic collocation method for nonlinear random vibration
Speaker: Jun He
Abstract: Based on Smolyak-type sparse grids, a stochastic collocation method is developed for the solution of nonlinear partial differential equations with high-dimensional random inputs (coefficients and forcing terms). The method does not suffer from the curse of dimensionality (computational cost exponentially growing with the number of random variables) and requires only repetitive runs of an existing deterministic solver to evaluate function values on a sparse grid. By setting the collocation point sets to be the cubature point sets obtained by using the chosen quadrature rules, numerical integration formulas are established for the evaluation of expectations of the solution. Numerical examples involving nonlinear elastic restoring force and hysteretic restoring force demonstrate the computational efficiency and accuracy of the method.

Date: Thursday March 8, 2012
Time: 3:30-5PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: The Bayesian world of engineering risk analysis
Speaker: Daniel Straub
Abstract: In the information age, Bayesian analysis is a key instrument in risk analysis for engineering systems. This talk aims to illustrate the potential of this instrument through a number of applications that we are currently researching in the ERA group at TU M�nchen. These applications require specific theoretical or algorithmic developments, some of which will be outlined in the talk. Applications to be presented include: Monitoring of geotechnical sites, emergency management for flood hazards, tunnel construction risk assessment, monitoring of deterioration concrete structures, wildfire risk assessment and water resource planning.

Date: Thursday March 1, 2012
Time: 3:30-5PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Modeling fatigue life of laminated composites using Bayesian inference
Speaker: Nikolay Dimitrov
Abstract: Fatigue of materials is a random phenomenon, usually associated with large scatter in properties. The large uncertainties pose difficulties in defining precise fatigue life models. Bayesian inference methods provide valuable tools for taking the uncertainties into account and defining fatigue models where the scatter in properties is realistically simulated. Due to the specifics of composites, estimating fatigue life of composite laminates requires an extra modeling step, where the fatigue properties of the entire stack of composite fabric layers (the laminate) are determined based on the properties of the elementary building blocks of material, the lamina. The present talk describes a model, estimating the fatigue properties of a laminated composite plate, based on material properties in the principal material directions of individual layers with unidirectional fibers (lamina). The model is tuned to measured data by the use of Bayesian inference, thus accounting for the inherent uncertainties.

Date: Thursday February 23, 2012
Time: 3:30-5PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Topics on uncertainty quantification and its challenges
Speaker: Khachik Sargsyan
Abstract: Over the past decade, considerable improvements in uncertainty quantification (UQ) algorithms, computational power, and data collection have advanced capabilities for UQ in predictive models to a significantly higher level. However, the current state-of-the-art techniques still face many challenges such as nonlinear, bifurcative systems, tail probabilities, the curse of dimensionality, limited sampling abilities, intrinsic stochasticity, input correlations, just to name a few. The seminar will go over some of these challenges and algorithmic developments to resolve them, driven mainly by applications in stochastic chemical kinetics and climate modeling. The backbone of our studies is the polynomial chaos theory with the Bayesian probabilistic point of view to enable both forward and inverse UQ. While illustrating sample techniques on specific applications, the presentation will hopefully trigger more general discussions on UQ methods.

Date: Thursday February 16, 2012
Time: 3:30-5PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Stochastic process representation: in the time or in the frequency domain?
Speaker: Marco Broccardo
Abstract: This talk presents a modern review of techniques used in representing stochastic processes. First, we discuss a time-domain representation of filter white noise, and a technique to reduce the number of random variables. Next, we present the frequency-domain formulation, discussing its advantages and limitations in detail. And lastly, we introduce the concept of Hilbert vector space and show how the two formulations can be revisited in this framework.

Date: Thursday February 9, 2012
Time: 3:30-5PM
Place: 542 Davis Hall
Topic: How should stochastic finite elements be set up to reflect nature?
Speaker: Martin Ostoja-Starzewski
Abstract: A random field is a prerequisite to solving a boundary value problem in mechanics of random media. The randomness in material properties is typically described by random fields with discontinuous realizations, such as they appear in image analysis of microstructures. However, smoothing on some intermediate scales (corresponding to a finite element size) is necessary to formulate mesoscale random fields, which then enter the stochastic finite element schemes. We discuss various consequences (non-uniqueness, scale-dependence, and anisotropy) imposed by physics and mechanics on those fields.

Date: Thursday February 2, 2012
Time: 3:30-5PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Bayesian network for infrastructure seismic risk assessment and decision
Speaker: Armen Der Kiureghian
Abstract: Bayesian network models for spatially distributed infrastructure systems subjected to earthquake ground motions will be described. The model consists of three parts: (a) a model of the earthquake hazard, accounting for uncertainties in the earthquake characteristics and location, as well as the spatial correlation structure of the ground motion random field, (b) models of the system components in terms of fragility functions, and (c) a model of the system performance in terms of component states. Considering the immediate aftermath of a major earthquake, the BN updating tool is used to assess the states of the infrastructure system and its components, as the available information about the hazard and from observed states of selected system components evolves in time. The updated probabilistic information, together with utility values for different outcomes, are used to make decisions regarding the operation of the system and its components, e.g., continued operation, operation at reduced capacity, or closure, or to select optimal ordering of inspections. A preliminary application of the model to a hypothetical model of the proposed California High Speed Rail system will be demonstrated.

Date: Thursday January 26, 2012
Time: 3:30-5PM
Place: 542 Davis
Topic: Stochastic model and simulation of near-fault ground motions for specified earthquake source and site characteristics
Speaker: Mayssa Dabaghi
Abstract: Near-fault ground motions often possess distinct characteristics, which can have strong influence on structural response. These include the rupture directivity effect in the fault-normal direction and the fling step in the fault-parallel direction. Due to scarcity of recorded near-fault ground motions, there is interest in developing synthetic ground motions which can be used in performance-based earthquake engineering in addition to or in place of recorded motions. It is crucial that such synthetic motions be realistic and have characteristics that are consistent with recorded near-fault ground motions. Furthermore, it is most useful if the synthetic motions are generated in terms of information that is normally available to the design engineer, such as the earthquake magnitude, the position of the site relative to the potential fault rupture, and the shear-wave velocity of the soil at the site.
In this presentation we develop a parametrized stochastic model of near-fault ground motion in the strike-normal direction. By fitting the model to a database of near-fault ground motions, we develop predictive equations for the model parameters in terms of the earthquake source and site characteristics mentioned above. Using these predictive equations, for a given set of earthquake and site characteristics, we generate sets of model parameters and use them to generate an ensemble of synthetic near-fault ground motions. The resulting synthetic motions have the same statistical characteristics as the motions in the database, including the variability for the given set of earthquake and site characteristics.

Date: Thursday January 19, 2012
Time: 3:30-5PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Response spectrum analysis for floor acceleration
Speaker: Matteo Pozzi
Abstract: Floor acceleration is a relevant engineering demand parameter in the design of buildings, as failure of non-structural components constitutes a significant portion of economic losses due to earthquakes. This presentation explores the possibility of adopting CQC (complete quadratic combination) method to estimate the maximum acceleration, as it is commonly used in the analysis of other parameters such as inter-story drift. The presentation will include discussion of the accuracy of the estimation, the effect of approximating the total acceleration with the pseudo-acceleration, and that of truncating the contribution of highest modes.

Date: Wednesday December 14, 2011
Time: 2-3:30PM
Place: 542 Davis Hall
Topic: Extreme loads on wind turbines
Speaker: Nikolay Dimitrov
Abstract: Estimation of the highest loads expected to occur during the lifetime of a structure is a common engineering problem. Lifetimes of structures are usually measured in decades, and we rarely have historical data covering such long periods. It is therefore necessary to make extrapolative predictions, where we try to estimate loads with recurrence periods considerably greater than the time period for which data are available. Due to the random nature of loads, this problem is often approached in a probabilistic manner. This presentation will try to facilitate a discussion on the statistical load extrapolation techniques and their application within the wind turbine design.

Date: Wednesday December 7, 2011
Time: 2-3:30PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Oil production in the North Sea from 1966 to 2011
Speaker: Wiggo Smeby
Abstract: Norway found oil in the North Sea sometime in the early 1960's and has since had a tremendous development in oil and gas production, including a significant amount of engineering challenges in the process. The presentation will focus on these challenges, and also give some indications of future energy and environmental developments.

Date: Wednesday November 30, 2011
Time: 2-3:30PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Nonlinear stochastic dynamic analysis in frequency domain using tail-equivalent linearization
Speaker: Marco Broccardo
Abstract: Analysis of structures subjected to stochastic excitation often requires consideration of nonlinear behavior. Furthermore, for highly reliably structures, interest is focused on the statistics of the response in the tail region of the underlying distribution. Recently, Fujimura and Der Kiureghian (2008) developed the Tail-Equivalent Linearization Method (TELM) to solve this class of problems. The approach, which is based on the First-Order Reliability Method (FORM), defines an equivalent linear system for each considered response threshold by equating its tail probability with the first-order approximation of the tail probability of the nonlinear response. Once the tail-equivalent linear system (TELS) is defined for a specific threshold, methods of linear random vibration analysis are used to compute various response statistics, such as mean crossing rates and tail probabilities of the local and extreme peak distributions. The original version of TELM was developed in time domain, and a first extension in frequency domain was developed by Garre and Der Kiureghian in 2010. In this talk, the general framework underlying TELM is first introduced and then followed by presentation of several recent developments for the frequency domain TELM.

Date: Wednesday November 16, 2011
Time: 2-3:30PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Applications of probabilistic design technologies in the design office: How the state of practice relates to the state of art
Speaker: Mohamed Talaat
Abstract: Probabilistic design methodologies have been applied by structural engineers for decades in the nuclear power industry, and more recently in performance-based seismic design of commercial buildings. Expectedly, the nature of regulated industry standards and its requirement for time-tested methods lags behind the frontiers of research. This seminar will present a detailed overview of the procedures currently employed in both disciplines (nuclear and commercial) and underscore concerns about the fidelity and shortcomings of these procedures. Subsequently, the seminar participants are invited to engage in a discussion about how best to address these concerns in a practical and efficient framework that can be implemented in a competitive industry context.

Date: Wednesday November 2, 2011
Time: 2-3:30PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic:Reliability-based design of wind turbine blades
Speaker:Nikolay Dimitrov
Abstract: Wind turbines are the largest revolving structures in the world. The current state-of-the-art wind turbine blades reach length of more than 60m and weigh more than 10 tons. Blades are exclusively made of composite materials, typically glass or carbon fiber composites. Wind turbines are characterized with complex structural properties and behaviour due to the use of advanced non-homogeneous materials and due to the coupled interaction between aerodynamic, elastic and inertial forces arising during operation. Significant uncertainties are associated with the material properties: composites exhibit a multitude of failure modes on material level, and due to the prevalence of manual labour in the manufacturing there is large chance for occurrence of manufacturing defects. Similarly, there are uncertainties associated with the loads due to exposure to varying, sometimes severe external conditions (turbulent wind and waves) and the presence of a control system influencing the loads.
Reliability analysis can be a valuable tool for gaining more knowledge and having better control over the uncertainties in wind turbine blade design. The present study has the following objectives:
-- Create methodologies for efficient reliability assessment of composite materials and composite wind turbine components. The assessment will be based on real measurements, numerical models and probability theory.
-- Map the uncertainties in processes, materials and external conditions involved in the lifecycle of the composite structural components that have effect on the health and behaviour of the composite structure.
The presentation will show a short introduction to the specifics of wind turbine (and wind turbine blade) design, and an overview of the current status of the project, with the methods used, the lessons learned, the challenges experienced until now, and the plans for further work.

Date: Wednesday October 26, 2011
Time: 2-3:30PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic:Multi-hazard interdependent reliability assessment of infrastructure networks with multistate components
Speaker:Mohammad Javanbarg
Abstract: Multi-hazard approach represents a convenient way to address system reliability of a critical infrastructure. The objectives of the present study are twofold. First, we present a multi-hazard methodology for evaluation of the reliability associated with the multistate damage of infrastructures subject to various hazards. The hazards can be considered as s-dependent, s-independent or mutually exclusive. By simulating multi-hazard effect as common cause failure (CCF) groups, an efficient multistate two-terminal reliability method is employed to multi-hazard reliability analysis of an infrastructure network. Multistate two-terminal reliability at a certain demand level can be defined as the probability that the system capacity generated by multistate components is greater than or equal to a specified certain demand. The terminal-pair reliability of the network is then evaluated by performing an efficient Monte Carlo simulation as an alternative to exact reliability calculation methods. Second, we present a novel methodology for interdependency analysis among the complex networks. By simulating the interdependencies among different networks using common cause failure and probability theory, the interdependent reliability of an infrastructure network can be calculated. A computer package has been developed based on the abovementioned methodologies namely UILLIS (Urban Infrastructure and Lifeline Interactions of Systems). A case study on a theoretical lifeline network is performed to illustrate the advantage of the package to multi-hazard interdependent reliability evaluation of a real-life network with multistate components.

Date: Wednesday October 19, 2011
Time: 2-3:30PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic:Quasi meshless discretization of random fields based on the Karhunen-Loeve expansion
Speaker:Wolfgang Betz
Abstract: The KL-expansion is a method for discretization of random fields, which is optimal among series expansion methods with respect to the mean square truncation error. The expansion is defined in terms of the eigenvalues and eigenfunctions of an integral eigenvalue problem with the autocovariance function as the integral kernel. Analytical solutions of this problem exist only for a few autocovariance functions and simple geometries. For the general case, the integral eigenvalue problem needs to be solved numerically, e.g. by application of the finite element (FE) method. However, FE methods require the construction of an FE-mesh, which may be a nontrivial task for domains with complex geometry. This talk introduces an alternative approach for the solution of the KL eigenvalue problem by application of the finite cell method. This method is based on embedding the actual domain in a larger domain defined by rectangular sub-domains, called finite cells. Accurate representation of the geometry is dealt with through the integration procedure, which is part of the FE solution process. In addition, the method uses high-order shape functions for the approximation of the solution inside the cells. This new approach is compared to the EOLE method, which is an alternative meshless method for the discretization of random fields in domains with arbitrary shape.

Date: Wednesday October 12, 2011
Time: 2-3:30PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Non-geometrical spectral characteristics for dynamic responses of linear structures to non-stationary random earthquake loads
Speaker: Jun He
Abstract: A numerical method in the time domain is developed to estimate the useful response spectral characteristics for the first-passage reliability problem of classically damped linear elastic structures to nonstationary random earthquake loads. Two random earthquake load models based on the definition of the evolutionary spectrum, i.e., the Conte-Peng model and Yeh-Wen model, are considered and reformulated in the present study. The unit impulse response functions (IRF) of the filters modeling ground motion acceleration and the whole linear systems subjected to white noise processes are derived, from which the time domain expressions of the response spectral characteristics under study are given out. Numerical investigations of the response spectral characteristics of a single-degree-of-freedom (SDOF) structure shows the efficiency of the developed method and reveals that the structural responses to the fully nonstationary earthquake load models are usually wide-banded, even for a SDOF structure. The seismic reliability estimate of a super-tall TV Tower demonstrates the use of the developed method in the firs-passage reliability problem.

Date: Wednesday October 5, 2011
Time: 2-3:30PM
Place: 542 Davis Hall
Topic: Non parametric regression for estimating the Value of Information
Speaker: Matteo Pozzi
Abstract: Structural Health Monitoring (SHM) systems are useful instruments for limiting the risk related to seismic events. The selection of these systems can be guided quantitatively by the Value of Information (VoI) principle of decision theory, predicting the reduction of overall losses. This seminar will introduce a numerical approach for estimating the VoI based on Monte Carlo simulation and non-parametric regression. Application of the proposed technique is presented for a bridge model subject to seismic excitation and instrumented with accelerometers.

Date: Wednesday September 21, 2011
Time: 2-3:30PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Efficient simulation of turbulent wind fields by auto-regressive vector processes
Speaker: Steen Krenk
Abstract: There are many advantages in simulating stochastic fields for e.g. wind, waves and earthquake motion by sequential simulation e.g. with the Auto-Regressive Moving Average (ARMA) format. However, the calibration of short filter representations have often constituted a problem. In the case of a turbulent wind field the covariance properties of the turbulent velocity components are well represented in a sufficiently large section transverse to the mean wind direction. This enables the use of a very short memory and thereby leads to very efficient simulation. The one-step autoregressive procedure is discussed in detail, and an explicit calibration procedure based on two sections separated by a fairly large number of time steps is presented. The method is applied to a fully three-dimensional convected isotropic turbulent wind field, and it is demonstrated that this simple simulation procedure reproduces the longitudinal and transverse correlation functions quite well, when calibrated by use of suitably separated sections. The method is potentially important in connection with large-scale wind turbines where the rotor diameter, and thereby the size of the relevant cross-section, is of the same order of magnitude as the correlation length of the wind field, i.e., around 100 m.

Date: Wednesday September 14, 2011
Time: 2-3:30PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Direct Differentiation Method (DDM) in the context of Tail Equivalent Linearization Method (TELM)
Speaker: Marco Broccardo
Abstract: The goal of TELM is to determine the statistics of the response of highly reliable, nonlinear systems. Based on the First-Order Reliability Method (FORM), TELM requires the implementation of an optimization algorithm to find the design point. Gradient-based optimization methods used for this purpose use the gradient of a limit state function and, in turn, that of the response. DDM is a robust and efficient method for computing the response gradient. Of course the response gradient can also be used in other applications, such as optimal design. This talk will discuss the implementation and testing of DDM for a Bouc-Wen oscillator.

Date: Wednesday September 7, 2011
Time: 2:30-3:30PM
Place: 542 Davis Hall
Topic: Influence of Frequency Distribution on Dynamic Response to Pedestrian Loads
Speaker: Steen Krenk
Abstract: The load of individual pedestrians has traditionally been modeled as occurring with a well defined 'walking frequency' that has then been assessed conservatively. In reality, the footfall frequency of the individual pedestrian exhibits a variability of about 5 pct. This implies a bandwidth of the load process that is larger than the spectral width of the resonance peak of the structure - typically in the order of 1 pct. It will be shown how the response to a fairly narrow-band process, representing the pedestrian load, can be obtained in closed form. The result combines the bandwidth of the load process and the damping ratio of the structure and typically leads to a reduction of the response by a factor of 0.2-0.3.

Date: Wednesday August 31, 2011
Time: 2-3:30PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: ARMENIA 2011 images from my memory
Speaker: Marco Broccardo
Abstract: Collections of landscapes, architecture, pieces of normal life and people grasped from a summer experience in Armenia will be presented. Sincere thanks goes to Professor Der Kiureghian, who gave me the opportunity not only to visit but also to live in this fascinating land.

Date: Wednesday May 18, 2011
Time: 1:3-3PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Overview of earthquake induced pipe damage
Speaker: Arthur Lebee
Abstract: Earthquakes strongly affect lifeline systems. Concerning water distribution system, everyone has in mind the consequences of the fire after the 1906 San Francisco earthquake. A good and fast estimate of pipe damage would be a useful tool for pre-event mitigation measures and for targeting repair teams after an event. This necessitates the derivation of a fragility curve relating raw seismic data and pipe failure.
Many attempt have been made to build such curve. In this overview, a description of the mechanisms relating earthquake motion to pipe failure is provided. Many open questions will be raised and are targeted to open discussion.

Date: Wednesday May 11, 2011
Time: 3-4:30PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Overview of earthquake induced pipe damage
Speaker: Mayssa Dabaghi
Abstract: A procedure for stochastically simulating an ensemble of near-fault strong ground motion time histories for a specified set of earthquake and site characteristics (earthquake magnitude, distance to rupture, shear wave velocity, location of the site relative to the fault, etc.) is presented. In this seminar, an example simulation of pulse-like motions is presented. Also, so far, a gamma modulating function has been used in modeling the residual. This form does not always seem appropriate, and a new formulation of the modulating function is proposed and investigated.

Date: Wednesday April 27, 2011
Time: 1:30-3PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: The utility of sensors for post-earthquake decision support
Speaker: Matteo Pozzi
Abstract: Immediately after a significant seismic event, the decision maker may rely on maps, estimating the ground motion and the shaking intensity. However, additional data collected by sensors and monitoring systems in real-time can better guide towards the optimal decision on a specific structure, avoiding mistakes and wastes of resources. This seminar investigates the utility of installing sensors on bridges for post-earthquake decision support, by estimating the "Value of Information" provided by these devises. Computational approaches based on discretization, policy evaluation and function approximation will be discussed.

Date: Wednesday April 20, 2011
Time: 1:30-3PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Stochastic geometric analysis of precipitates in a material grain
Speaker: Armen Der Kiureghian
Abstract: The yield strength of certain material grains depends on the shapes, sizes and distribution of precipitate intersections with the (1,1,1) plane of the grain. Analytical expressions are derived for the statistics of these quantities for cubical and spherical precipitates with constant and variables sizes.

Date: Wednesday April 6, 2011
Time: 1:30-3PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Bayesian methods in engineering risk analysis
Speaker: Daniel Straub
Abstract: The talk will present an overview on the application of Bayesian methods in structural reliability and risk analysis. Selected aspects of the theory will be highlighted, including the use of Bayesian networks, and its application to a number of relevant problems will be shown, including modeling of inspection in deteriorating systems and natural hazards management. The talk will conclude with a discussion of future research needs.

Date: Wednesday March 30, 2011
Time: 1:30-3PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Engineering characterization of spatially variable earthquake ground motions
Speaker: Timothy Ancheta
Abstract: Earthquake ground motions exhibit spatial variability manifest as random variations of Fourier amplitude and phase. These variations increase with frequency and distance between observations points (x), and can introduce significant demand for lifeline systems (e.g., pipelines) and foundations. Spatially variable ground motions (SVGM) are quantified by: (1) apparent horizontal wave velocity (Vapp), which controls wave passage effects that shift Fourier phase; (2) lagged coherency, representing random phase variations; and (3) standard deviation terms representing Fourier amplitude variability. A simulation procedure is developed that modifies a seed motion in a manner compatible with the three sources of SVGM. An application of simulated SVGM is illustrated. Suites of SVGMs are generated and strain histories computed. Peak ground strains (PGS) are found to increase with peak velocity (PGV), similar to previous work, but also decrease with x and saturate for PGV >~80 cm/s, which has not been previously observed. The dependence of PGS on x is confirmed from array recordings.

Date: Wednesday March 16, 2011
Time: 1:30-3PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Stochastic Estimate of Ground Motion Derived from Aftershock Recordings: the example of Oceano, California, for the M 6.5 22 December 2003 San Simeon Earthquake
Speaker: Carola Di Alessandro
Abstract: The U.S. Geological Survey deployed a digital seismic station in Oceano, California, in February 2004, to investigate the cause of damage and liquefaction from the 22 December 2003 M 6.5 San Simeon earthquake. We analyzed the recorded aftershocks together with recordings of the mainshock and the same aftershocks obtained from nearby stations, to estimate the mainshock ground motion in Oceano. We tested a set of aftershocks as Green's functions by comparing simulated and recorded acceleration amplitude spectra for the mainshock at nearby stations. A stochastic operator was convolved with the aftershock accelerograms so to simulate the duration and phase of the mainshock accelerograms. This approximation allowed us to extend the range of aftershocks that can be used as Green's functions to events nearly three magnitude units smaller than the mainshock.

Date: Wednesday March 9, 2011
Time: 1:30-3PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: A review of the use of Bayesian networks in health care
Speaker: Iris Tien
Abstract: Bayesian networks (BNs) are a graphical representation of random variables and their probabilistic interdependencies. They have been used in many applications to analyze systems and support decision making. This seminar will review the state of the art in using BNs in health care, including: what medical applications it has been used in, the strengths and benefits of adopting this approach, and the potential challenges of implementing BNs as the framework for future systems. Studies on using BNs in the processing of fMRI data, in monitoring health outcomes, for large diagnostic systems, and related to the value of information in health risk management will be presented.

Date: Wednesday March 2, 2011
Time: 1:30-3PM
Place: 542 Davis
Topic: Model-based and residual-based approaches to damage detection
Speaker: Matteo Pozzi
Abstract: The Bayesian approach allows to process rationally the measurements collected on civil structures for damage detection, by using a predictive model. However, its implementation can be computationally expensive. In this seminar, the method is presented, and compared with alternative techniques based on the analysis of residuals, which are model-free and cheaper to implement.

Date: Wednesday February 23, 2011
Time: 1:30-3PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Stochastic simulation of near-fault ground motions for specified earthquake and site characteristics
Speaker: Mayssa Dabaghi
Abstract: A procedure for stochastically simulating an ensemble of near-fault strong ground motion time- histories for a specified set of earthquake and site characteristics (earthquake magnitude, distance to rupture, shear wave velocity, location of the site relative to the fault, etc.) is presented. In particular, one can generate an ensemble of synthetic near-fault ground motions in the fault normal direction, having or not a forward directivity pulse. Use of simulated motions is of particular interest in performance-based earthquake engineering due to scarcity of recorded near-fault ground motions. The simulated motions can be used in addition to or in place of recorded motions.

Date: Wednesday February 16, 2011
Time: 1:30-3PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Bayesian network for infrastructure seismic risk assessment and decision support
Speaker: Armen Der Kiureghian
Abstract: Bayesian networks provide an excellent framework for seismic infrastructure risk assessment and decision-support, particularly in near-real time applications when it is necessary to make emergency decisions in light of observations made during a major earthquake. This talk will present the development and application of such a framework for a hypothetical model of the proposed California high speed rail system.

Date: Wednesday February 9, 2011
Time: 1:30-3PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Damping scaling factor for elastic response spectra
Speaker: Sanaz Rezaeian
Abstract: Ground motion prediction equations (GMPEs) are often used in structural engineering to predict the response spectral values for a given set of earthquake and site characteristics. The Next Generation Attenuation (NGA) models are a set of five GMPEs that were developed in 2008 and are widely used in the current practice. These models predict the elastic response spectra for 5% viscous damping. For many structural engineering applications, a more accurate estimate of the damping ratio is required. The current guidelines to translate the 5% damped elastic response spectra to other damping ratios pre-date NGA and are outdated. In this project, we develop a new model for a scaling factor that can be used to adjust the 5% damped spectral ordinates to spectral ordinates for other damping ratios. This is done empirically, using a newly developed database of over 6000 records. Damping ratios from 0.5% to 30% are considered.

Date: Wednesday February 2, 2011
Time: 1:30-3PM
Place: 542 Davis Hall
Topic: On frequency-domain simulation of synthetic ground motions for specified earthquake and site characteristics
Speaker: Marco Broccardo
Abstract: In this talk we will review the basic approach to the frequency-domain extension of the stochastic ground motion model and simulation method developed by Sanaz Rezaeian and Armen Der Kiureghian. The method employs an evolutionary power-spectral density based on the Clough-Penzien-Kanai-Tajimi filter with time-varying parameters. As in the time-domain approach, the temporal and spectral nonstationary characteristics of the model are completely separate. Furthermore, the modulating function is identical to that used in the original model and its parameters are the same. However the power spectral density is characterized by 4, not 2 parameters. The challenge will be to relate these two new parametric variables with physical properties of ground motions. Once the global framework of this extension is completed, we will be able to simulate synthetic ground motions with a remarkable reduction in the number of standard normal random variables utilized for discretization of the process. This aspect is relevant in computational methods for non-linear random vibration analysis, such as TELM.

Date: Wednesday January 26, 2011
Time: 1:30-3PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: How to assess the value of information provided by a monitoring system
Speaker: Matteo Pozzi
Abstract: In Structural Health Monitoring, tests and sensing systems are intended as tools providing diagnoses, which allow the operator of the facility to develop an efficient maintenance plan or to require extraordinary measures on a structure. The relevance of these systems depends directly on their capability to guide towards the most optimal decision for the prevailing circumstances, avoiding mistakes and wastes of resources. In this context, the concept of "Value of Information" provides a rational approach to rank measuring systems according to a utility-based metric, which fully includes the decision making process affected by the campaign. This framework allows, for example, an explicit assessment of the economical convenience of adopting a sensor depending on its precision. The seminar will introduce the basic concepts, provide applications to reliability and maintenance problems, and discuss specific topics such as alternative computational approaches and the adoption of non-linear cost vs utility mapping.

Date: Wednesday December 8, 2010
Time: 3:30-5PM
Place: 542 Davis
Topic: Thick periodic plates homogenization, application to sandwich panels including chevron folded core
Speaker: Arthur Lebee
Abstract: Sandwich panels are widespread in everyday life. Their structural efficiency is well-known and is a central criterion in possible applications. This presentation is dedicated to the study of a new sandwich panel core which might replace honeycomb in some applications: the chevron pattern. In order to compare this new core to other ones, an accurate knowledge of its mechanical behavior is necessary. However, the price for structural efficiency is a more complex mechanical behavior. This is the case for the shear forces stiffness which is critical when comparing sandwich panels cores. Thus, in this work we reconsider in details and in the general case how to derive plates behavior under shear forces. A new plate theory is suggested as well as the related homogenization scheme for periodic plates. This plate theory is the extension of the well-known Reissner-Mindlin plate theory in the case of heterogeneous plates. However, it cannot be reduced to a Reissner-Mindlin plate theory in the general case. In the special case of sandwich panels including the chevron pattern, applying the homogenization scheme brings out a skins distortion phenomenon which affects a lot their shear forces stiffness.

Date: Wednesday November 10, 2010
Time: 3:30-5PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Infrastructure Interdependency Modeling with I2Sim
Speaker: Smitha Koduru
Abstract: Interdependencies among modern infrastructure systems are often complex and are frequently overlooked at a system level decision-making. In the post-disaster management, focusing on any infrastructure system without recognizing potential interdependencies would hinder the restoration of functionality. The Joint Infrastructure Independency Research Program (JIIRP) at the University of British Columbia (UBC) employs systems engineering approach to address the issue of interdependencies and coordination among multiple infrastructure systems to achieve community resilience. The system of infrastructure systems with static and dynamic interdependencies, and dynamic islanding for the prevention of cascading failures, are simulated with Infrastructure Interdependency Simulator (I2Sim). An overview of JIIRP and the methodology of I2Sim are presented in this talk.

Date: Wednesday November 3, 2010
Time: 3:30-5PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Stochastic simulation of near-fault ground motions for specified earthquake and site characteristics
Speaker: Mayssa Dabaghi
Abstract: A procedure for stochastically simulating an ensemble of near-fault strong ground motion time- histories for a specified set of earthquake and site characteristics (earthquake magnitude, distance to rupture, shear wave velocity, location of the site relative to the fault, etc.) is presented. Use of simulated motions is of particular interest in performance-based earthquake engineering (PBEE) due to scarcity of recorded near-fault ground motions. The simulated motions can be used in addition to or in place of recorded motions.

Date: Wednesday October 27, 2010
Time: 3:30-5PM
Place: 542 Davis Hall
Topic: Reliability-based design of slope angle considering spatial variability of soil material
Speaker: Iason Papaioannou
Abstract: In this talk, we present a methodology for reliability-based design of slopes, taking into account the spatial variability of the strength parameters of the soil material. The objective is evaluation of the optimal slope angle for a given reliability constraint related to the stability of the slope. The corresponding performance function is expressed in terms of the factor of safety, computed by applying a strength reduction technique to a 2-dimensional elasto-plastic finite element model. The optimal slope angle is determined using an inverse reliability optimization algorithm, derived by modifying the Hasofer-Lind-Rackwitz-Fiessler method used for the evaluation of the most probable failure point in the first order reliability method, where the search space is extended to include the design parameter. The spatial variability of the uncertain strength parameters is represented by a non-Gaussian random field. The underlying Gaussian field is discretized applying an embedded domain approach, whereby the actual domain is embedded in a standard volume and the Karhunen-Loeve expansion is applied to the latter. Therefore, the random field discretization remains unchanged throughout the optimization procedure, not being affected by the change of the geometry of the soil profile. A numerical example demonstrates the solution approach and investigates the importance of the spatial variability of the soil material on the optimal slope angle.

Date: Wednesday October 13, 2010
Time: 3:30-5PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: A portable time domain electromagnetic array for identification of unexploded ordnance
Speaker: Karl N. Kappler
Abstract: A significant problem exists both internationally and domestically regarding vast tracts of land contaminated by buried unexploded ordnance. An estimated 10M acres of land in the United States alone is rendered unusable due to presence UXO. I will present an overview of the technology past, present, and future which is used to locate ordnance and discriminate geophysical anomalies from clutter. The vast majority holes which are dug in co-location with geophysical anomalies are 'dry', i.e. they contain only harmless scrap metal. The Berkeley UXO Discriminator (BUD) system is a prototype device designed for simultaneous acquisition of multiple-channel measurements, thus enabling array style processing from single occupations of a given location. The BUD system has been field tested at several live sites with encouraging results. Key to BUD's success is the consonance of onboard hardware, firmware, and software in order to process field data in real time. Field data are inverted for best fit object position, and a set of three 'polarizability curves' which succinctly summarize the multiple channel data to give the operator information about the shape and symmetry of the target being interrogated. Current research with polarizability curves utilizing feature-based pattern recognition techniques has led to a scheme called 'Polarizability Curve Matching' which successfully identifies the class of buried object without false negatives.

Date: Wednesday October 6, 2010
Time: 3:30-5PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Frequency-domain simulation of synthetic ground motions for specified earthquake and site characteristics
Speaker: Marco Broccardo
Abstract: A frequency-domain extension of the stochastic ground motion model and simulation method recently developed by Sanaz Rezaeian and Armen Der Kiureghian is presented. The method employs an evolutionary power-spectral density based on the Clough-Penzien-Kanai-Tajimi filter with time-varying parameters. As in the time-domain approach, the temporal and spectral nonstationary characteristics of the model are completely separate. Furthermore, the modulating function is identical to that used in the original model and its parameters are the same. Hence, it is only necessary to additionally estimate the parameters of the filter by statistical analysis of observed records. The advantage of this extension is the remarkable reduction in the number of standard normal random variables utilized for discretization of the process. This aspect is relevant in computational methods for non-linear random vibration analysis, such as TELM.

Date: Wednesday September 1, 2010
Time: 3:30-5PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Feasibility of FORM for PBEE applications
Speaker: Smitha Koduru
Abstract: This presentation focuses on two broad themes; first, treatment of uncertainties and the application of the first-order reliability method (FORM) in the finite-element reliability analysis, and second, the seismic risk assessment of reinforced concrete structures for performance states such as, collapse, and monetary loss. Topics related to these areas, such as, second-order approximation for probability content in the standard normal space, and effects of spectral nonstationarity on structural damage, are also discussed.

Date: Wednesday July 21, 2010
Time: 2-2:45PM
Place: 542 Davis
Topic: High dimensional model representation -- applications in probabilistic analysis
Speaker: Devdas Menon
Abstract: This presentation aims to show how "High Dimensional Model Representation (HDMR)" can be used as a computationally efficient tool in mapping the input-output relationship in complex systems. HDMR can be used to express the response function in terms of a summation of component functions, involving hierarchical correlated function expansions. The application of HDMR and its efficiency are demonstrated with reference to (1) solving some standard reliability problems, (2) developing seismic fragility curves (of an RC building), and (3) carrying out global sensitivity analysis (of a meso-scale model of an unreinforced masonry shear wall).

Date: Wednesday July 21, 2010
Time: 3-3:45PM
Place: 542 Davis
Topic: Bayesian updating for long-term monitoring of structures
Speaker: Matteo Pozzi
Abstract: Novel sensing technologies, as fiber optic-based strain gauges, today offer durable solutions for structural monitoring. However, the measurements recorded have to be rationally interpreted, taking the relevant uncertainties into account. This talk presents the basics and some applications of the Bayesian framework to accomplish this task.

Date: Wednesday May 26, 2010
Time: 3:30-5:00PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Influence of site classification on computing empirical ground-motion prediction equations in Italy
Speaker: Carola Di Alessandro
Abstract: In this study, we investigate a site classification method for stations of the Italian Accelerometric Network based on the predominant period of ground motion at the site. The site predominant period is identified from the average horizontal-to-vertical (H/V) spectral ratios of the 5%-damped response spectra of Italian earthquake records. Whenever possible, we classified sites by assigning them to one of six predominant period classes (in the range 0.05 to 2 seconds) that we propose as a modification of the Zhao et al. (2006) procedure. We then investigated the impact of this classification scheme on empirical ground-motion prediction equations. We adopted the same functional form of Fukushima et al. (2007) and we computed a nonlinear period-dependent regression that allowed us to derive site coefficients using the proposed six predominant period classes. Our empirical site classification scheme based on strong-motion data provides the opportunity to explore whether we can decrease the misfit by improving the site characterization of the Italian data set. Comparison of our results with other empirical ground-motion prediction equations (GMPEs) based on conventional site classifications do not display a significant reduction of overall standard deviation. However, our site classification schemes shows promise in reducing the uncertainty in ground-motion prediction. The advantage over the simpler classification schemes used in many GMPEs is that we capture the effect of deep and shallow soil profiles and high shear-wave velocities. Another advantage of our classification scheme is that it is based on a relatively quick and inexpensive measurements and interpretation.

Date: Wednesday May 19, 2010
Time: 3:30-5:00PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: A simplified Bayesian network for post-earthquake decision on monitored structures
Speaker: Sara Broglio
Abstract: In every urban grid it is possible to identify structures with strategic function (e.g. hospitals, police and fire stations, highways/railways bridges). For such structures it is important to assess the condition of and make decisions immediately after an earthquake. Nowadays monitoring structures with accelerometers is increasingly feasible. A simplified Bayesian Network (BN) framework for processing the evolving information that becomes available from the sensors placed in the structure, but also from seismological sources or recordings of the ground motion in nearby sites that become available shortly after the earthquake, is presented with an example for a typical building. The probabilistic information coming from the BN associated with information on the costs associated with various states of teh building are then used to formulate and solve an optimal decision problem.

Date: Wednesday April 28, 2010
Time: 3:30-5:00PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Estimation of coherency for an array of near-fault records
Speaker: Katerina Konakli
Abstract: In previous seminars, we examined the effect of the earthquake spatial variability on structural response as well as a method of simulating differential support motions conditioned on actual records. In both cases, the ground motion spatial variability is modeled in the frequency domain through the complex valued coherency function. The modulus of the coherency function characterizes the incoherence effect, i.e. the loss of coherency of seismic waves with distance due to reflections and refractions and also due to the differential super-positioning of waves arriving from different parts of an extended source. This effect is described mathematically either by theoretical models based on the physics of the phenomenon or by empirical functions fitted to observed data. In this seminar, we analyze the coherency characteristics of the UPSAR near-fault array of motions recorded during the 2004 Parkfield Earthquake of magnitude 6.0. The coherency estimates from the three components of motion recorded at 12 stations are discussed and compared with existing theoretical and empirical models.

Date: Wednesday April 21, 2010
Time: 3:30-5:00PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: A discussion about near-source seismic demand: The case of l'Aquila earthquake
Speaker: Eugenio Chioccarelli
Abstract: Rupture directivity effects in ground motion are known since many years to both seismologists and earthquake engineers; i.e., in sites that are in a particular geometrical configuration with respect to the rupture, the velocity fault-normal signals may show a large pulse which occurs at the beginning of the record and contains the most of energy. The results are waveforms different from ordinary ground motions recorded in the far field or in geometrical conditions not favorable with respect to directivity. During the seminar increments in both elastic and inelastic seismic actions quantified using the American Next Generation Project (NGA) database will be presented and results will be confronted with the characteristics of the mainshock (Mw 6.3) of the recent Abruzzi (central Italy) sequence that is the third strongest seismic event producing strong motion records in Italy, after the Irpinia (1980, Mw 6.9) and Friuli (1976, Mw 6.4) earthquakes. The analyses employ recently developed tools and they represent one of the first attempts to identify near-source effects in the data recorded during a normal faulting earthquake.

Date: Wednesday April 14, 2010
Time: 3:30-5:00PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Bayesian network modeling of system performance
Speaker: Michelle Bensi
Abstract: Bayesian networks provide an excellent framework for modeling system performance, particularly in near-real time applications when it is necessary to update models in light of available observations. However, Bayesian networks can be very demanding of computer memory and inference can become intractable if care is not taken to optimize their topology. In this paper, efficient BN formulations for modeling system performance are presented. First formulations are developed for series and parallel systems and then results are extended to general systems for which the minimal link sets and/or minimal cut sets are known. An optimization algorithm is presented to automate the generation of efficient BN formulations for modeling system performance. An example application demonstrates the computational efficiency of the system model obtained by solving the optimization problem.

Date: Wednesday March 31, 2010
Time: 3:30-5:00PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Signal Processing and Probabilistic Seismic Hazard Analysis Tools for Characterizing Near-Fault Directivity
Speaker: Jack Baker
Abstract: Near-fault ground motions containing strong velocity pulses are of interest to engineers designing systems close to active faults. These ground motions have been identified as imposing extreme demands on structures, but methods of accounting for their effects in design are still relatively ad hoc. This presentation reviews a proposed quantitative ground motion classification scheme that is being utilized to identify these ground motions, study their unique effects on structures, and account for their occurrence in probabilistic seismic hazard analysis. Example results are shown to illustrate the classification scheme, how it can be used to develop statistical models that predict the occurrence and severity of this effect in future ground motions, and what impact these models have on probabilistic seismic hazard analysis calculations. A proposed strategy for selection of recorded ground motions that properly represent potential directivity effects at a site is presented.

Date: Wednesday March 17, 2010
Time: 3:30-5:00PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: On the optimal discretization of Gaussian random fields
Speaker: Iason Papaioannou
Abstract: This presentation will focus on an attempt to optimize the performance of the EOLE method (Expansion Optimal Linear Estimation) in the discretization of Gaussian random fields. The EOLE method is based on a point discretization scheme, namely the OLE method (Optimal Linear Estimation), whereby the dimensionality is reduced by performing a spectral decomposition of the covariance matrix corresponding to the nodal values of the random field. Alternatively, the dimensionality reduction could be achieved by the solution of an optimization problem, requiring the minimization of the error in the resulting approximation of the covariance matrix. We investigate the performance of both approaches against different error measures, including the error in the reliability estimation.

Date: Monday March 15, 2010
Time: 3:30-5:00PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: A Bayesian Network Framework for Seismic Infrastructure Risk Assessment & Decision Support
Speaker: Michelle Bensi
Abstract: A Bayesian network (BN) is a probabilistic graphical model that represents a set of random variables and their probabilistic dependencies. The variables may represent demand or capacity values, or the states of components and systems. BNs are graphical and intuitive, facilitate information updating, can be used for identification of critical components within a system, and can be extended by decision and utility nodes to solve decision problems. In particular, the facility for information updating renders the BN an ideal tool for infrastructure risk assessment and decision support. Evidence on one or more variables (e.g. observed component capacities, demands, or states) can be entered into the BN and this information propagates throughout the network to provide an up-to-date probabilistic characterization of the performance of an infrastructure system under an evolving state of information.

This presentation will begin with a brief introduction to BNs. Next, a broad overview of a BN framework for infrastructure seismic risk assessment and decision support will be presented. Components of the framework include: (1) a seismic demand model of ground motion intensity as a spatially distributed Gaussian random field accounting for finite fault rupture and directivity effects, (2) models of component performance, (3) models of system performance, and (4) the extension of the BN to include decision and utility nodes to aid post-earthquake decision-making. Like all computational methods, BNs have limitations. In particular, calculations in BNs can be highly demanding of computer memory. A discussion of work done to address this limitation will be presented. An illustrative example will demonstrate the main ideas and approach. The presentation will conclude with a broader discussion of the use of BNs for modeling hazards, processing information, and aiding decision making with the goal of improving infrastructure system performance and sustainability.

Date: Wednesday March 10, 2010
Time: 3:30-5:00PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Reliability Updating with Equality Information
Speaker: Daniel Straub
Abstract: In many instances, information on engineering systems can be obtained through measurements, monitoring or direct observations of system performances and can be used to update the system reliability estimate. In structural reliability analysis, the information is expressed either by inequalities (e.g. for the observation that no defect is present) or by equalities (e.g. for quantitative measurements of system characteristics). In the latter case, hitherto, common structural reliability methods (SRM) were not directly applicable to the computation of the updated reliability. In this presentation, a novel representation of equality information is introduced, which enables the use of SRM for this computation. The approach will be demonstrated on numerical examples, including an application to fatigue reliability and corrosion measurements.

Date: Wednesday March 3, 2010
Time: 3:30-5:00PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Modeling of Near-Fault Ground Motion Directivity Pulses
Speaker: Mayssa Dabaghi
Abstract: The difference in the characteristics of near-fault ground motions with forward directivity pulses and far-field ground motions has resulted in a growing interest in studying the characteristics of these special types of ground motions and investigating their effects on structures. Near-fault forward-directivity motions should be considered in design. However, records are scarce and modern codes do not represent them adequately. Therefore need arises to characterize and simulate such ground motions. Several authors have already investigated near-fault ground motions and have proposed mathematical models to represent them or to modify response spectra such that they account for directivity effects. This presentation will go over a proposed methodology based on existing literature to extract and model directivity pulses for the ultimate purpose of simulating synthetic near-fault ground motions for a specified set of earthquake and site characteristics. The pulse extraction procedure proposed by Jack Baker is being applied to identify and extract the pulses from the near-fault records of the recent L'Aquila earthquake (Mw=6.3), in order to investigate their spatial variability. Some preliminary results are presented. The Masters thesis "Characterization of Near-Fault Forward-Directivity Ground Motions" of Suparno Mukhopadhyay is also reviewed, since it falls within the same scope of work.

Date: Wednesday February 17, 2010
Time: 3:30-5:00PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Multi-Peril Loss Estimation Without Double Counting
Speaker: Petros Keshishian
Abstract: Catastrophe risk modeling usually deals with the derivation of the probability distribution of the damage to the given asset or portfolio of assets due to a given hazard, where damage represents the monetary payout to restore the assets to their pre-damaged condition. In case of a single peril, like earthquake ground shaking, the estimation of the damage to the given asset comes down to the modeling of its vulnerability - that is the level of the damage given the intensity of the ground shaking. But there are cases when a single event can cause multi-peril damage to the given asset. For example, earthquake ground shaking and fire following, hurricane wind and flooding, tornado wind and hail, etc. In such cases, the definition of the damage as a payout to restore the asset to its pre-damaged condition still holds, except the payout should include the total payout required to restore the damage from all perils without double counting. Double counting refers to the overlap of the estimation of damage from multiple perils, e.g. fire damage to a building that has been lready destroyed due to earthquake ground shaking. The latter is known as the problem of "burning the ruble". This presentation will outline a comprehensive approach for modeling the combined damage from multiple perils without double counting. The developed framework derives the mean and the full probability distribution function of the combined damage for multiple perils. Consideration is given to cases where damage from one peril could impact the vulnerability of the asset to subsequent perils.

Date: Wednesday February 3, 2010
Time: 3:30-5:00PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Effect of random field approximation on system failure probability
Speaker: Michelle Bensi
Abstract: We are currently developing a Bayesian network (BN) framework for seismic infrastructure risk assessment and decision support. Infrastructure systems exposed to geographically distributed hazards (e.g. earthquakes) require that loads be modeled as a spatially distributed random field. However, modeling correlated random variables drawn from a random field via BN is computationally demanding and can become intractable for large infrastructure systems. To overcome this problem, we have developed approximation methods to capture as much of the correlation structure as possible, without sacrificing computational tractability. In this seminar we investigate the effect of these approximations on the computed system failure probability.

Date: Wednesday January 27, 2010
Time: 3:30-5:00PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Stochastic simulation of ground motion components with correlated parameters
Speaker: Sanaz Rezaeian
Abstract: A method for simulating orthogonal horizontal components of ground motion with correlated parameters for specified earthquake and site characteristics is presented. Correlations between parameters (e.g., Arias intensity, effective duration, frequency, etc.) of the two ground motion components are determined empirically by analyzing a large database of recorded motions. Given earthquake and site characteristics (i.e., faulting mechanism, moment magnitude, source-to-site distance, and soil type), correlated parameters for the two ground motion components are randomly generated. A stochastic model is then utilized to simulate the two components along the principal axes of ground motion, along which the two components are statistically independent for the given parameters. The simulated components can then be rotated into any desired direction (e.g., principal directions of the structure) through a simple transformation. This method is currently applicable to far-field motions with a source-to-site distance of at least 10km. Accounting for correlations between parameters of ground motion components is important in order to obtain synthetic motions that are realistic. This is especially important when analyzing asymmetric structures that are vulnerable to torsion. In practice, these correlations are often neglected, but as shown in this presentation they can be very high and should be accounted for.

Date: Wednesday January 20, 2010
Time: 3:30-5:00PM
Place: 542 Davis Hall
Topic: Random field reliability analysis in geotechnical engineering by application of subset simulation
Speaker: Iason Papaioannou
Abstract: In geotechnical engineering, the uncertainties stem to a large extent from the stochastic properties of the materials within the soil mass. The spatial variability of each stochastic parameter is represented by a random field. The discretization of the random fields usually leads to a large number of random variables. This presentation will focus on the efficient reliability assessement of such high dimensional problems by applying an adaptive Monte Carlo method, known as subset simulation. This method enables the efficient estimation of the failure probability, by expressing the failure event as the intersection of intermediate failure events, corresponding to larger conditional probabilities. The intermediate probabilities are estimated by applying an efficient Markov chain Monte Carlo technique. The method overcomes the inefficiency of the Monte Carlo simulation in estimating small probabilities, while maintaining its independency on the problem dimensionality.

Date: Thursday December 17, 2009
Time: 3:30-4:30PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: A review of the characteristics and proposed models of near-fault ground motions
Speaker: Mayssa Dabaghi
Abstract: The difference in the characteristics of near-fault ground motions with forward directivity pulses and far-field ground motions has resulted in a growing interest in studying the characteristics of these special types of ground motions and investigating their effects on structures. Moreover, modern codes do not represent near-fault motions adequately. Several authors have investigated near-fault ground motions and have proposed mathematical models to represent them or to modify response spectra such that they account for directivity effects. This presentation will go over the existing literature concerning near-fault ground motions and present and compare the findings of various authors. Of interest is to ultimately develop a more comprehensive model of near-fault ground motions to be able to generate realistic earthquake ground motions that are consistent with recorded near-fault motions with forward directivity.

Date: Thursday December 10, 2009
Time: 3:30-4:30PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Development of a BN for seismic risk assessment of the proposed California High Speed Rail: Preliminary Models and Results
Speaker: Michelle Bensi
Abstract: We are currently working on the development of a Bayesian Network model for seismic risk assessment of the proposed California High Speed Rail system. The focus of this seminar will be on model formulations and assumptions. Critique of these formulations/assumptions (and alternative suggestions) will be solicited. Preliminary results based on the current assumptions will be presented.

Date: Thursday December 3, 2009
Time: 3:30-4:30PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Enhanced Bayesian network for structural and system reliability assessment and decision
Speaker: Armen Der Kiureghian
Abstract: Bayesian network (BN) and structural reliability methods (SRMs) are combined to develop an enhanced BN (eBN) that allows consideration of continuous nodes with arbitrary distributions. Through a process of node elimination, a reduced BN (rBN) is derived that only has discrete nodes and can be solved with existing exact inference algorithms. SRMs are used to compute the conditional probability tables in the rBN. The rBN allows efficient life-cycle analysis and decision of structural and infrastructure systems.

Date: Thursday November 19, 2009
Time: 3:30-4:30PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Computationally tractable Bayesian network modeling of random fields
Speaker: Michelle Bensi
Abstract: This presentation will describe a recent investigation into Bayesian network (BN)-based methods for modeling spatially-distributed random fields. An introduction to random fields will be presented followed by a discussion of approaches for modeling random fields via BN. Approximation methods are discussed for ensuring computationally tractable BN-based random field models. The results of a numerical investigation of several generic but systematic spatial configurations are presented to compare the effectiveness of the proposed approximation methods. Time-permitting, an example application of modeling ground motion intensity as a spatially distributed random for a proposed high-speed rail system will be presented.

Date: Thurdasy November 12, 2009
Time: 3:30-4:30PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Simulation of spatially varying ground motions
Speaker: Katerina Konakli
Abstract: Time-history analyses of bridges for spatially varying earthquake input are typically performed with simulated ground motions, since closely spaced recorded motions are rare. The Conditional Probability Density Function (CPDF) method can be used to generate a random ground motion at a specified location compatible with predefined time-histories at separate stations and a coherency function. In this seminar, we describe the CPDF method and present example sets of support motions generated by conditioning on actual earthquake records for different levels of incoherency. The non-stationarity of the earthquake motions is accounted for by dividing the original records into nearly stationary segments. The required post-processing of the simulated time-histories is also described.

SPECIAL EVENT: Reliability Symposium: Presentations by students in CE229 -- Structural and System Reliability. See program.
Date: Tuesday, May 19, 2009
Time: 9:10AM - 3:00PM
Place: 406 Davis Hall

Date: Monday, May 4, 2009
Time: 2:00-3:30PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic:Variations on MCFM for nonlinear stochastic dynamic analysis
Speaker:Umberto Alibrandi
Abstract: It has been shown recently that FORM can effectively solve certain nonlinear stochastic dynamics problems. A major issue in this approach is determination of the design point. An adaptive version of the Model Correction Factor Method (MCFM) will be presented for solving this problem.

Date: Monday, April 27, 2009
Time: 2:00-3:30PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Modeling infrastructure systems with multi-state components and system state using BN
Speaker: Michelle Bensi
Abstract: In this seminar, I will recap the binary system connectivity BN formulations presented in previous seminars. Then, I will present BN formulations that can accommodate components with multiple damage states (e.g. slight, moderate, extensive, complete) or capacities (e.g. full, half, zero) and the resulting multi-state system. The max-flow min-cut theorem will be utilized to achieve this goal.

Date: Monday, April 13, 2009
Time: 2:00-3:30PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Random-effects regression modeling for analysis of clustered data and its application in ground motion simulation
Speaker: Sanaz Rezaeian
Abstract: Stochastic simulation of a suite of ground motion time-series for a specific design scenario requires development of predictive relations for the parameters of the stochastic ground motion model based on earthquake and site characteristics. To predict such relations, collected data from previous earthquake recordings are analyzed. For each earthquake in the database there are various numbers of recordings at different locations and sites. It is reasonable to assume that the data corresponding to the ground motions that come from the same earthquake are correlated, while the data corresponding to different earthquakes are statistically independent. In other words, the data is clustered. Unlike ordinary regression analysis, random-effects regression modeling assumes that data within clusters are dependent to some degree.
In this seminar, random-effect regression modeling is used to develop predictive relations for the stochastic ground motion model parameters based on earthquake and site characteristics. The validity of the regression formulas are examined by analysis of residuals. Finally, simulated ground motions based on these new relations are compared to the NGA ground motion prediction models in terms of spectral values.

Date: Monday, April 6, 2009
Time: 2:00-3:30PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Investigation of the fling step and its effect on bridge response
Speaker: Katerina Konakli
Abstract: The slip-parallel component of a near-fault earthquake motion is characterized by a permanent ground displacement, known as the fling step. However, the fling step does not appear in the earthquake records of the NGA database, which have been filtered so that the ground displacements recover to zero. In this seminar, we study the characteristics of the fling step in association with the earthquake kinematics and dynamics and demonstrate how it can be added to the NGA ground motions. Furthermore, we examine the effect of the fling step on structural response and explore methods to account for it in the Multiple Support Response Spectrum Analysis (MSRS) analysis of bridges.

Date: Monday, March 16, 2009
Time: 2:00-3:30PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: A gradient-free method for application of FORM to nonlinear stochastic dynamic analysis
Speaker: Umberto Alibrandi
Abstract: It has been shown recently that FORM can effectively solve certain problems in nonlinear stochastic dynamics. In this talk an adaptive version of the Model Correction Factor Method (MCFM) is presented, which allows determining the design point efficiently and without computing the response gradient.

Date: Monday, March 9, 2009
Time: 2:00-3:30PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Enhanced Bayesian networks: theory and applications
Speaker: Daniel Straub
Abstract: The newly developed enhanced Bayesian network (eBN) methodology combines Bayesian networks with structural reliability methods. In this presentation, the theory of eBN will be summarized and illustrated by means of examples, which include the reliability analyses of a structural system and of a portfolio of structural systems.

Date: Monday, March 2, 2009
Time: 2:00-3:30PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Further studies on the MSRS method for seismic analysis of bridges
Speaker: Katerina Konakli
Abstract: This seminar will focus on two main issues within the framework of the Multiple Support Response Spetcrum (MSRS) analysis. The first is the MSRS analysis of bridges crossing faults. The MSRS estimates of the peak responses of an example bridge structure are compared with results from time-history analyses for three recorded near-fault ground motions and various orientations of the bridge relative to the fault. The second issue investigated is the effect of modal truncation in the MSRS analysis. The errors resulting from mode truncation are analyzed by examining separately the static modal responses and the dynamic modal amplifications. The results for the cases of uniform and variable excitations are compared.

Date: Monday, February 23, 2009
Time: 2:00-3:30PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Updated Bayesian Network framework for modeling seismic demands and infrastructure system performance
Speaker: Michelle Bensi
Abstract: This presentation will provide an overview of our recent efforts towards the development of a Bayesian Network-based probabilistic decision-support system (DSS) for seismic infrastructure risk management (with a focus on near-real time emergency response and recovery following an earthquake). We model seismic demands on an infrastructure system by constructing a BN model of ground motion intensity as a spatially distributed random field. In this presentation I will briefly present this seismic demand model with recent enhancements including the addition of a finite source model and framework for inclusion of directivity effects. Finally I will present a BN model for infrastructure system performance assessment using fragility functions and five BN system connectivity formulations.

Date: Monday, February 9, 2009
Time: 2:00-3:30PM
Place: PEER Center Conference Room (325 Davis Hall)
Topic: Simulation of ground motions for given earthquake and site characteristics
Speaker: Sanaz Rezaeian
Abstract: Current ground motion prediction models (a.k.a. attenuation laws) are designed to predict measures of the ground motion intensity associated with a given set of earthquake and site characteristics. Such attenuation models are useful for response spectrum analysis, but they do not predict time-histories of future earthquakes, and with the advent of nonlinear dynamic analysis there is an increased need for prediction of �site-specific� acceleration time-histories. The main purpose of this seminar is to demonstrate how an ensemble of simulated ground motions can be generated for a given set of earthquake and site characteristics. Whenever appropriate, our formulas will be compared to the existing empirical relations.
In this seminar, a stochastic model that adequately captures the temporal and spectral nonstationarities of earthquakes is used. Identification of the model parameters for a large database of strong ground motion records provides the necessary observational data to perform statistical analysis such as distribution fitting and regression. Appropriate distribution functions are assigned to each model parameter, and after a transformation to the standard normal space, linear regression is used to construct relations between the transformed model parameters and the earthquake and site characteristics. The resulting formulas are used to randomly generate a set of correlated model parameters for a specific site of interest. Using these randomly generated parameters in the stochastic model, suites of strong ground motions are simulated that can replace actually recorded earthquakes (for example in PBEE analysis).

Date: Thursday December 18, 2008
Time: 3:30-4:30 PM
Place: PEER Center Conference Room (325 Davis Hall)
Title: Hurricane-related Multi-hazard Risk Analysis
Speaker: Ning Lin, Princeton University
Abstract: Hurricanes constitute one of the major hazards for the eastern United States. Associated with strong winds, heavy rainfall, and storm surge, landfalling hurricanes often cause enormous structural damage and economic losses to coastal regions. A deterministic and probabilistic framework to assess the risk of hurricane-induced damage is proposed. Three of the four main components of the framework are discussed in detail. We first present a structural vulnerability model that couples an innovative windborne debris risk module and a component-based pressure damage module to estimate the accumulative wind damage to residential developments during a hurricane passage. Second, the simulation of landfalling hurricanes using the Weather Research and Forecasting (WRF) model, with initial and boundary conditions generated by the Geophysical Fluid Dynamics Laboratory�s (GFDL) hurricane model, is introduced, with as an example the simulation of Hurricane Isabel (2003). Third, we illustrate the storm surge simulation using the Advanced Circulation model (ADCIRC) for the New York Harbor region. In the end of the presentation, the fourth component of the analysis framework, also our current research focus, on hurricane climatology and landfalling probability estimation, is briefly discussed.

Date: Thursday December 4, 2008
Time: 3:30-4:30 PM
Place: PEER Center Conference Room (325 Davis Hall)
Title: The Model Correction Factor Method in the Random Vibration Analysis
Speaker: Umberto Alibrandi
Abstract: Random vibration analysis aims at determining the response statistics of linear and nonlinear systems, when subjected to a stochastic excitations. Recently, it has been shown that certain problems in random vibrations can be solved using the computational tools (FORM/SORM) of the time-invariant structural reliability theory. However they require the determination of the design point, given as the solution of a constrained optimization problem; if the limit state surface is nonlinear the optimization problem must be solved by an iterative scheme. Using a gradient-based approach, this typically involves repeated calculations of the limit state function and its gradient with respect to selected realizations of the random variables, which implies several deterministic dynamic analyses. In order to reduce the computational effort, the adoption of the Model Correction Factor Method (MCFM) is proposed in this talk. It is a type of Response Surface approach, where the FORM is applied to a simple model (here chosen as linear), and then a model correction factor is introduced to move the linearization point of the idealized problem onto the limit-state surface of the original problem. In this way an upper bound of the reliability index is obtained, which can be corrected by using the uniform directional simulation on a cone. This technique allows determining the correction angle of the linear idealized problem, requiring a sample size much smaller than the dimension of the space of the basic variables. The application of the method to simple non-linear oscillators shows the computational efficiency of the proposed procedure.

Date: Thursday November 20, 2008
Time: 3:30-4:30 PM
Place: PEER Center Conference Room (325 Davis Hall)
Title: The Multiple Support Response Spectrum method: overview, improvements and example applications
Speaker: Katerina Konakli
Abstract: The Multiple Support Response Spectrum method is used to investigate the effect of spatial variability of earthquake ground motion on the response of Caltran�s bridges. The important practical issue of the required number of modes in the MSRS analysis is discussed. Furthermore, the improvement achieved with the extended MSRS method that accounts for high-frequency modes is assessed. Finally, the problem of bridges crossing faults is considered. The seismic demands for various bridge orientations are estimated in an example application. This application demonstrates the ability of the MSRS method to give good approximations of the peak response for a case of strongly non-stationary and narrow-band excitation.

Date: Thursday November 13, 2008
Time: 3:30-4:30 PM
Place: PEER Center Conference Room (325 Davis Hall)
Title: Optimization Under Uncertainty for Families of Systems and Systems of Systems
Speaker: Mark Philip McDonald, Asst Prof. of Civil Engineering, Vanderbilt University
Abstract: Decision making and design for families of systems (FoS) and systems of systems (SoS) problems is difficult because of complex interactions among member systems, uncertainties in the problem environment, and the presence of multiple decision makers. This presentation develops a computational framework for decision making in an SoS context based on the techniques of reliability-based multidisciplinary design optimization (RBMDO) and stochastic programming. This approach is chosen because these techniques allow decision makers to model system interactions, propagate uncertainties, and analyze decisions in an integrated method. This research pursues the following objectives: (1) develop single-loop formulations for performing reliability-based design optimization (RBDO) with component and system reliability constraints and algorithms which are more efficient than standard nonlinear programming methods for solution, (2) develop methods for performing RBDO with continuous and discrete design and random variables, (3) extend RBDO techniques developed in objectives 1 and 2 to FoS problems; (4) apply RBMDO techniques to SoS problems with cooperative decision makers who delegate decision making to a central authority using systems dynamics modeling of the system of systems and RBMDO methods for system of systems optimization under uncertainty, and (5) apply methods for optimization under uncertainty to SoS problems where there are autonomous decision makers via agent based modeling and stochastic programming.The proposed methods are illustrated with example problems from structural optimization, aerospace vehicle sizing, network allocation, emergency response, and optimization of traffic controls.

Date: Thursday November 6, 2008
Time: 3:30-4:30 PM
Place: PEER Center Conference Room (325 Davis Hall)
Title: Modeling seismic demands and system performance of a spatially distributed infrastructure system using a Bayesian network
Speaker: Michelle Bensi
Abstract: In previous seminars, the application of Bayesian Networks (BNs) to seismic infrastructure risk management (and associated challenges) was presented. In this seminar, I will present a revised BN model for modeling seismic demands on a spatially distributed infrastructure system. Additionally, I will discuss how we can model system performance using a BN framework. Finally, I will present a "live" demonstration of the BN model (with examples of "information updating") as applied to a simple transportation system using the Hugin computer program.

Date: Monday October 27, 2008
Time: 3:30-4:30 PM
Place: PEER Center Conference Room (325 Davis Hall)
Title: Tail equivalent linearization method in presence of materials with asymmetric behavior
Speaker: Salvatore Sessa
Abstract: The conventional linearization methods used in random vibration, require the assumption of a distribution for the nonlinear response to allow computations of the second moments and usually a Gaussian distribution is employed. The Tail Equivalent Linearization Method (TELM) has been recently developed in order to evaluate the non-Gaussian response in time of a non linear system by matching the tail probability of the linearized system with the first order approximation of the non-linear tail probability of the system. In case of asymmetric hysteresis loop of the non linear material, this feature of the TELM becomes more interesting since TELM is able to catch the asymmetry of the response probability distribution between positive and negative side. In this seminar, the application of the TELM to an asymmetric generalized Bouc-Wen oscillator will be shown; more in detail, there will be an introduction about the requirements of the constitutive model of the material and its properties, then the description of the TELM algorithm in presence of asymmetry. Finally, a two degrees of freedom asymmetric oscillator will be analyzed in order to get the IRFs of the linearized system, the response-in-time probabilistic characterization and the distribution of the extreme response in case of stationary excitation.

Date: Thursday, October 16, 2008
Time: 3:30-4:30 PM
Place: PEER Center Conference Room (325 Davis Hall)
Title: Structuring complex systems using Bayesian Networks
Speaker: Peter Friis Hansen and Michelle Bensi
Abstract: In the presentation we will first give a basic introduction to the Bayesian Networks. We will make use of the program "Hugin" to present illustrative modeling of selected Bayesian network. We will model a few simple reliability problems and discuss how these results compare to standard FORM/SORM results. Modeling principles and techniques will be discussed and illustrated in the presentation of selected larger worked examples.

Date: Thursday, October 9, 2008
Time: 3:30-4:30 PM
Place: PEER Center Conference Room (325 Davis Hall)
Title: An Alternative Response Surface Methodology
Speaker: Umberto Alibrandi
Abstract: In order to reduce the computational costs in structural reliability analysis of complex systems it has been suggested to use Response Surface Methods (RSM). The basic idea is to replace the original performance function by a simple and approximate explicit function, named Response Surface (RS), whose function values can be computed more easily. However, it has been shown that the usually adopted Quadratic RS (QRS) may give results which are very sensitive to the experimental plan. This drawback is overcome by the Support Vector Method (SVM), which guarantees good robustness properties, but not fast convergence towards the exact limit state surface. An alternative RSM, called Kernel Response Surface (KRS), is proposed in this talk. KRS provides a trade-off solution between the QRS and the SVM. Some simple numerical examples show the accuracy and effectiveness of the proposed procedure.

Date: Thursday, October 2, 2008
Time: 3:30-4:30 PM
Place: PEER Center Conference Room (325 Davis Hall)
Title: Estimating the response of long period structures with closely spaced modes to near-field ground motion
Speaker: Jyothi Krishnan
Abstract: Response of long period structures very close to earthquake faults is currently estimated using the CQC modal combination rule. Near-fault motions are highly non-stationary (of shorter duration compared to structure period and containing pulses of particular frequencies). This violates the basic assumptions behind the CQC rule. In this seminar a procedure to estimate response to near-field ground motion as a combination of response to an extracted pulse and to the residual broad-band motion is presented. The method will then be applied to two structures designed such that their response is not modeled well by the CQC rule.

Date: Thursday, September 25, 2008
Time: 3:30-4:30 PM
Place: PEER Center Conference Room (325 Davis Hall)
Title: Modeling the quality of inspections for spatially distributed defects
Speaker: Daniel Straub
Abstract: A quantitative inspection model describes the ability of an inspection method to identify and characterize defects. Available inspection models, such as Probability of Detection (PoD), assume that the potential location of the defect is known. For many deterioration mechanisms in engineering systems, notably localized corrosion in pressure vessels and pipelines, the assumption does not hold. This presentation introduces a model of inspection quality that relaxes this assumption, based on a simplified spatial model of defect distribution. The model is validated numerically, using simulated defects. To facilitate reliability updating following an inspection, a generic limit state function for inspection outcomes is introduced. Finally, application of the model in reliability updating is demonstrated, considering inspections of deteriorating pipelines.

Date: Thursday, September 11, 2008
Time: 3:30-4:30 PM
Place: 542 Davis Hall
Title: Development of Future Aircraft Structural Engineers: Today's Cutting Edge, Tomorrow's Old School
Speaker: Michael Mohaghegh
Abstract: Advanced materials, designs, and computational methods used on development of future aircraft require structural engineers that have the latest competencie in these areas. Be prepared to stay ahead of the curve in your career. Come to hear about the multiple programs offered at Boeing to develop world class aircraft structural engineers working in a cntinuous learning environment.

Date: Thursday, September 4, 2008
Time: 3:30-4:30 PM
Place: PEER Center Conference Room (325 Davis Hall)
Title: Seismic Response Analysis of a Bridge Crossing a Fault
Speaker: Katerina Konakli
Abstract: My study of structural response to spatially varying ground motions, in this seminar, is specialized on bridges crossing fault rupture zones. After a brief discussion of the characteristics of near-fault seismic motions, we will look into the response of an existing bridge designed by Caltrans. The ability of the Multiple Support Response Spectrum (MSRS) method to predict the structural response for the specific case of spatial variability is assessed. The analysis method proposed by Chopra and Goel (2008) for bridges crossing faults is also discussed.

Date: Thursday, August 28, 2008
Time: 3:30-4:30 PM
Place: PEER Center Conference Room (325 Davis Hall)
Title: Likelihood estimation of parameters using simultaneously monitored processes
Speaker: Peter Friis-Hansen
Abstract: The topic of my talk in the reliability seminar is on maximum likelihood inference from several simultaneously monitored response processes of a structure to obtain knowledge about the parameters of other not monitored but important response processes when the structure is subject to some Gaussian load field in space and time. The considered example is a ship sailing with a given speed through a Gaussian wave field.

Date: Thursday, August 14, 2008
Time: 3:30-4:30 PM
Place: PEER Center Conference Room (325 Davis Hall)
Title: Application of the Bucket Elimination algorithm to a BN for seismic infrastructure risk management
Speaker: Michelle Bensi
Abstract: In previous seminars, the application of Bayesian networks to seismic infrastructure risk management and the associated challenges were discussed. One of these challenges involved the lack of computational efficiency and tractability associated with many of the available inference algorithms. Without an efficient inference algorithm, the proposed model would be rendered useless. In this seminar, the Bucket Elimination algorithm, and the associated computational complexity, will be presented. In addition to providing an explanation of the algorithm, time will be spent discussing why and how this algorithm can be applied to the proposed BN.

Date: Tuesday, July 29, 2008
Time: 3:30-4:30 PM
Place: PEER Center Conference Room (325 Davis Hall)
Title: Generating artificial ground motions for use in PBEE
Speaker: Sanaz Rezaeian
Abstract: A nonstationary stochastic model of strong ground motions, discussed in previous seminars, is used to simulate accelerograms, given only a set of earthquake and site characteristics (i.e. type of faulting, magnitude, distance from fault, and soil conditions), without any need for a previously recorded earthquake in the region. Such a model facilitates generating a suit of artificial ground motions that is appropriate for use in performance-based earthquake engineering (PBEE). The focus of this seminar is to investigate relations that will help us to predict the model parameters for a given set of earthquake and site characteristics. To develop such predictive relations, model parameters are identified for a data-set of strong ground motion recordings, statistical data analysis is performed, and some preliminary results, including fitted distributions of the model parameters and the correlations between parameters are presented.

Date: Tuesday, June 24, 2008
Time: 3:30-4:30 PM
Place: PEER Center Conference Room (325 Davis Hall)
Title: Berkeley revisited
Speaker: Daniel Straub
Abstract: I will look back at my last two years spent at Berkeley and review my main findings and some trivia. To end my postdoc time on a sweet note, for once this presentation will contain more pictures than equations and I will also provide some dessert.

Date: Tuesday, June 10, 2008
Time: 3:30-4:30 PM
Place: PEER Center Conference Room (325 Davis Hall)
Title: Multiple support response spectrum analysis of bridges III
Speaker: Katerina Konakli
Abstract: In this seminar we will look into the effect of ground motion spatial variability on an existing bridge structure designed by Caltrans. The efficiency of the improved MSRS method that accounts for high-frequency modes will be discussed. Finally, we will focus on the site-response effect of the coherency function and compare two different models for the frequency-response function of a soil column.

Date: Tuesday, June 3, 2008
Time: 3:30-4:30 PM
Place: PEER Center Conference Room (325 Davis Hall)
Title: TELM at sea: an application to an offshore structure - part II
Speaker: Luca Garre
Abstract: In this talk I will address the derivation of an alternative TELM model defined in the frequency domain. The new Tail Equivalent Linearized System (TELS) is defined through a Frequency Response Function (FRF), which is directly computed in terms of the gradient of the limit state function. An analytical derivation of the FRF will be presented, together with a more intuitive geometrical approach. I will present some first results with attention to invariance of the FRF with respect to the scaling and, to a lesser extent, the spectral shape of the excitation.

Date: Tuesday, May 27, 2008
Time: 4-5 PM
Place: PEER Center Conference Room (325 Davis Hall)
Title: Enhancing Seismic Hazard Assessment with Machine Learning: Applications of Bayesian Networks and Support Vector Machines
Speaker: Michelle Bensi
Abstract: Machine learning is concerned with the development and use of algorithms that allow computers to �learn.� That is, the algorithms improve automatically (based on computational and statistical methods) through experience. Machine learning encompasses a wide range of topics and this seminar will focus on two of these topics: Bayesian Networks and Support Vector Machines (SVMs). This will be a two part seminar. The first portion will focus on the application of Bayesian Networks to the spatial modeling of earthquake hazard and infrastructure system performance (recently presented at: Inaugural International Conference of the Engineering Mechanics Institute). The second portion will focus on prediction of ground motion intensity (specifically whether intensity exceeds a specified threshold) using SVMs and logistic regression. In addition to providing a brief background on SVMs, results from SVMs and logistic regression will be compared with predictions from two attenuation relations.

Date: Tuesday, May 13, 2008
Time: 4-5 PM
Place: PEER Center Conference Room (325 Davis Hall)
Title: Stochastic modeling of deterioration processes through dynamic Bayesian networks
Speaker: Daniel Straub
Abstract: A generic framework for stochastic modeling of deterioration processes is proposed, based on dynamic Bayesian networks (DBN), which can be interpreted as a generalization of Markov chain models. The framework facilitates computationally efficient and robust reliability analysis and, in particular, Bayesian updating of the deterioration models with measurements, monitoring and inspection results. These properties make it ideally suited for near-real time applications in asset integrity management and deterioration control. The framework is demonstrated and investigated through two examples considering fatigue crack growth.

Date: Tuesday, May 6, 2008
Time: 4-5 PM
Place: PEER Center Conference Room (325 Davis Hall)
Title: Statistical approach for evaluation of identified vibration properties of a long-span bridge using a wireless sensor network
Speaker: Shamim Pakzad
Abstract: A wireless sensor network for structural monitoring applications is designed and developed with the goal of deployment of a scalable network with high data quality on a large infrastructure. The scalability of the network in terms of the number of nodes is provided by implementation of a novel pipelining technique for reliable dissemination of commands and data collection. The scalability in terms of the quality of the vibration data is provided by achieving time synchronization for sampling rates of up to 5 kHz, and sensitivity for a broad range of accelerations from low-level ambient vibrations to high-level earthquake motions. During the three months test deployment of the network on the Golden Gate Bridge, independent of the number of the nodes, a steady communication bandwidth of 550 bytes per second was maintained. Using the ambient vibration data collected by the network, the vibration modes of the bridge, including higher modes of up to 5 Hz frequency, with a high degree of confidence were identified. Distribution properties of the modes were estimated and a statistical platform for evaluating the accuracy of the identified parameters was used. This integrated approach to hardware architecture, system software, and statistical system identification is required for achieving wireless sensor networks with high spatial density and data quality that can be effectively utilized for health monitoring of infrastructures.

Date: Tuesday, April 29, 2008
Time: 4-5 PM
Place: PEER Center Conference Room (325 Davis Hall)
Title: FORM in high dimensions
Speaker: Armen Der Kiureghian
Abstract: Some researchers have recently criticized applications of the first-order reliability method to problems with large numbers of random variables. One such application area is nonlinear stochastic dynamic analysis by FORM, which requires discrete representation of the excitation by a large number of random variables. This issue will be revisited in this presentation with special focus on nonlinear stochastic dynamic analysis by the Tail-Equivalent Linearization Method, TELM.

Date: Tuesday, April 22, 2008
Time: 4-5 PM
Place: PEER Center Conference Room (325 Davis Hall)
Title: Attenuation of Inelastic Response Spectra
Speaker: Yousef Bozorgnia
Abstract: Practical seismic hazard analysis is almost universally carried out for peak ground motion values and linear elastic response spectra. During a severe ground shaking, however, many structures behave inelastically. In order to carry out a direct hazard analysis on other intensity measures (IMs) such as inelastic response spectra, attenuation models for the new IM have to be developed. In this study:
-- Inelastic response spectra for thousands of recorded ground motions compiled in PEER NGA database are computed,
-- Attenuation models for inelastic response spectral ordinates are developed, and
-- The inelastic attenuation models are used to perform direct probabilistic seismic hazard analysis (PSHA) on inelastic response spectra.
An advantage of this process is that no assumption is made about estimating inelastic response from elastic response spectra, as the new attenuation models are directly developed based on inelastic spectral ordinates. Also, the new attenuation models are easily implemented into a standard PSHA package--in this case, open-source OpenSHA software developed by the USGS.

Date: Tuesday, April 15, 2008
Time: 4-5 PM
Place: PEER Center Conference Room
Title: TELM at sea: an application of the Tail Equivalent Linearization Method to an offshore structure
Speaker: Luca Garre
Abstract: The talk will deal with random vibrations analysis of offshore structures. Focusing on a given offshore platform I will present the following topics:
-- Linear Airy theory for sea waves kinematics;
-- Structural and loading (Morison's equation) models employed in the analysis;
-- FORM analysis for the most probable excitation leading to exceedance of a given treshold at a given time;
-- Computational problems arising with the TELM procedure in connection with excitations defined in the frequency domain;
-- A possible approach to overcome the problems and some preliminary considerations.

Date: Tuesday, April 8, 2008
Time: 4-5 PM
Place: PEER Center Conference Room
Title: New developments on a stochastic ground motion model & its application in Earthquake Engineering
Speaker: Sanaz Rezaeian
Abstract: Stochastic ground motion models are needed in the current practice of performance-based earthquake engineering (PBEE). These models enable engineers to generate artificial samples of ground motions, which can be used in the place of actually recorded accelerograms when the scarcity of recorded earthquakes in a region causes difficulty in selecting an appropriate suite of ground motions.

Many stochastic ground motion models have been proposed, but the task of selecting an appropriate set of model parameters for a site of interest remains unresolved. The goal of this seminar is to discuss constructing distribution models for the stochastic model parameters and formulating correlations between these parameters and the earthquake and site characteristics by performing statistical analysis on a large number of recorded ground motions.

Date: Tuesday, April 1, 2008
Time: 4-5 PM
Place: 721A Davis Hall
Title: Information Updating in Infrastructure Systems Subject to Multiple Hazards
Speaker: Daniel Straub
Abstract: A main difficulty in managing infrastructure subject to multiple hazards is the large uncertainty associated with engineering models of the hazards, infrastructure response and the societal demand on the infrastructure. Any attempt to optimize hazard mitigation and recovery must account for these uncertainties, ideally through a probabilistic assessment. On the other hand, with the developments and increased availability of IT and measurement technology (e.g., remote sensors or GPS), more and more information on the hazards, the infrastructure condition and the system demand are readily available. Information reduces uncertainty in our models and in doing so supports the identification of optimal decisions. Therefore, there is a huge potential for computationally efficient methods for updating of the probabilistic models under new evidence. This presentation will present examples of how this is done and it will outline how these methods can be integrated into a new generation of decision support systems.

The presentation will start by discussing challenges involved in safeguarding infrastructure from multiple hazards. The Bayesian network methodology as a tool for jointly representing and analyzing multiple hazards will be introduced, and it will be demonstrated how this methodology can efficiently perform information updating under new evidence. The use of the methodology for optimizing decisions regarding hazard prevention, protection systems, early warning, inspection and control, as well as recovery, will be outlined. The concepts and mathematical tools will be illustrated by examples considering avalanche, earthquake, rock-fall and deterioration of infrastructures from the academic and industrial experience of the speaker.

Date: Tuesday, March 11, 2008
Time: 4-5 PM
Place: 721A Davis Hall
Title: Reliability-based demand analysis of structures subjected to ground shaking
Speaker: Marcello Bianchini
Abstract: The correct assessment of different performance levels of nonlinear modeled structures represents one of the most significant goals in the field of earthquake engineering. Starting from the expression for the classical time-independent reliability formulation, and under a few established assumptions, the probability of exceeding a specified performance level can be written in a closed form. The method represents a specific application of the Probabilistic Seismic Demand Analysis (PSDA), defined in the PEER loss estimation framework. For a given structure in a known seismic environment, the definition of which Intensity Measure (IM) plays a leading role in the whole process. In fact, the ground motion IM defines the seismic hazard at the site and it can be used as scale factor of records in incremental dynamic analysis to assess the fragility functions. A generic IM must satisfy some desirable properties in order to obtain reliable results. An average of spectral accelerations over a certain range of periods is first studied as a statistical predictor of nonlinear structural response of buildings subjected to recorded ground motions, and second compared with traditional elastic and inelastic scalar IMs. Finally, for a set of RC frame structures, which have been designed according to Eurocode 8 and capacity design criteria, PSDA is carried out combining the failure mechanisms, and the effectiveness of a well-selected IM is shown.

Date: Tuesday, March 4, 2008
Time: 4-5 PM
Place: 721A Davis Hall
Title: TELM analysis by FERUM and implementation of degrading materials
Speaker: Salvatore Sessa
Abstract: A new package for performing TELM analysis by FERUM will be presented together with an example application of an SODF oscillator used for testing the package. Furthermore, the seminar will present a generalized Bouc-Wen degrading material model and a first attempt at using it in conjunction with TELM.

Date: Tuesday, February 26, 2008
Time: 4-5 PM
Place: 721A Davis Hall
Title: More on the Application of Bayesian Networks to Seismic Hazard Assessment
Speaker: Michelle Bensi
Abstract: In a previous seminar, the application of graphical models, specifically Bayesian Networks, to seismic hazard assessment was discussed. That presentation focused particularly on the inclusion of spatial error correlation and how the addition of this correlation can be modeled efficiently. This seminar will continue with the discussion of these issues and expand upon the application of aforementioned ideas for actual modeling. An example of a simple spatially distributed system will be presented.

Date: Tuesday, February 12, 2008
Time: 4-5 PM
Place: 721A Davis Hall
Title: Multiple support response spectrum analysis of bridges
Speaker: Katerina Konakli
Abstract: Continuing the study of structural response to differential support motions, we use a more refined model that accounts for three-dimensional ground excitation and includes the site response effect in the coherency function. Furthermore, we discuss the question of how many modes need to be used in the analysis. Finally, we investigate the specific case of a bridge crossing the fault: the coherency function and the structural response are evaluated for various orientations of the bridge model.

Date: Tuesday, February 5, 2008
Time: 4-5 PM
Place: 721A Davis Hall
Title: Modeling deterioration and optimizing inspection efforts in spatially distributed systems - the example of corrosion in large RC structures
Speaker: Daniel Straub
Abstract: I will present an overview on the probabilistic modeling of corrosion of the reinforcement in concrete structures, with emphasis on the aspects of spatial variability. A hierarchical model is introduced for representing these spatial aspects. The model is then applied for optimizing inspection and maintenance efforts in large concrete structures. This presentation is based on earlier work (http://www.informaworld.com/10.1080/15732470601017369) but here the methods and results are discussed in light of Bayesian network models that we are currently working on.

Date: Tuesday, January 29, 2008
Time: 4-5 PM
Place: 721A Davis Hall
Title: The Model Correction Factor Method: an efficient response surface technique
Speaker: Luca Garre
Abstract: Structural reliability analysis (SRA) makes use of so-called Response Surface Methods (RSMs) for the solution of the reliability problem in the case of difficult-to-solve structural analyses. One of such analyses could be the assessment of buckling or ultimate resistance of bi-axially loaded steel stiffened plates, as it is the case in ship engineering practice. In these nonlinear cases, the computation of the structural response might be difficult, and integration of the structural code into the SRA algorithm is uneasy or simply impossible. Therefore, RSMs aim at defining an alternative limit state surface to be introduced into the SRA procedure by a sequence of repetitive calls to the structural code, according to a somehow defined grid of points in the space of the random variables.
In this presentation I will discuss about a particular response surface technique, the so- called Model Correction Factor Method (MCFM). According to this method, a simplified and easy-to-solve formulation of the problem at hand is employed as the limit state function. Such a code is then calibrated to a more advanced and realistic one (such as FEM) in a probabilistic sense, as the calibration is performed around the design point resulting from application of the SRA algorithm with the simplified limit state function. After a limited number of such adaptive calibrations the series of design points converges.
After a brief introduction to the procedural background of the method, I will show an application of this procedure to a typical marine problem, defined by a stiffened bottom plate subjected to both lateral and axial loads.

Date: Wednesday, December 12, 2007
Time: 12-1 PM
Place: 721A Davis Hall
Title: Application of Bayesian Networks to Seismic Hazard Assessment
Speaker: Michelle Bensi
Abstract: The assessment of seismic risk has become increasingly common in the field of earthquake engineering where there is interest in determining the hazard associated with a future earthquake of random characteristics (e.g. magnitude and location) at a site. This presentation will discuss the application of graphical models to seismic hazard analysis. Graphical models, specifically Bayesian Networks, are a favorable way to model seismic hazard because they allow for ease in updating the model as additional information becomes available.

Specifically, this presentation will focus on examining the importance of including spatial correlation of attenuation relation error terms when predicting spectral acceleration at a site using a graphical model approach. Additionally, this report will consider the effect of approximating this correlation by clustering correlation into geographically discrete segments in the spatial domain to increase computational efficiency. The goal of this report is to understand the �trade-off� between including spatial correlations versus the reduction in computational efficiency its inclusion creates.

Date: Wednesday, December 5, 2007
Time: 12-1 PM
Place: 721A Davis Hall
Title: Nonlinear Random Vibration Analysis by using TELM in OpenSees
Speaker: Sanaz Rezaeian, Salvatore Sessa
Abstract: In the previous seminar, we talked briefly about the reliability and TELM (tail equivalent linearization method) modules in OpenSees. In this seminar, we will use the results of the TELS (tail equivalent linear system) obtained from OpenSees to analyze the behavior of the system subjected to stationary and non-stationary excitations. Sample results for a SDOF system and a MDOF system (truss structure) will be presented.

Date: Wednesday, November 28, 2007
Time: 12-1 PM
Place: 721A Davis Hall
Title: The effect of structural redundancy on the acceptability of deterioration
Speaker: Daniel Straub
Abstract: Criteria for the acceptability of deterioration in structures are required to determine and document the appropriateness of mitigation actions such as design provisions, maintenance, retrofit and inspections. Because deterioration is modeled at the level of structural elements, these criteria must also be formulated at the element level, and, therefore, should account for the structural importance of the elements. However, for redundant structural systems, the quantification and application of such criteria has been hindered in the past by the difficulties involved in the reliability analyses of such systems. I present a method for establishing reliability acceptance criteria for deterioration in elements of redundant structural systems based on overall acceptance criteria on the system level. The proposed method utilizes equivalent structural systems that facilitate reliability analysis. These equivalent structural systems are established by calibrating them to characteristics of the real structure, which include a) the reliability of the structure without deterioration, b) indicators for the redundancy of the structure with respect to deterioration failure of individual elements, and c) estimates of the statistical dependence among deterioration at different locations in the structure. I will conclude the presentation by numerical examples demonstrating the effect of structural redundancy on the target reliabilities of structural elements.

Date: Wednesday, November 14, 2007
Time: 12-1 PM
Place: 721A Davis Hall
Title: Modal combination rules for response spectrum analysis of bridges
Speaker: Armen Der Kiureghian
Abstract: Several response spectrum methods for seismic analysis of bridges will be reviewed. Consideration will be given to multi-component support motions, interacting vector responses, and spatial variability.

Date: Wednesday, November 7, 2007
Time: 12-1 PM
Place: 721A Davis Hall
Title: Seismic Damage and Loss Models for Reinforced Concrete Bridges
Speaker: John-Michael Wong
Abstract: Post-earthquake repair costs are important for evaluating the performance of new bridge design options and existing bridge configurations. A step-by-step probabilistic repair cost methodology is proposed to evaluate the cost of repairs for bridge components and the bridge as a system for varying degrees of damage. This seminar will demonstrate an application of this methodology for evaluating column design options and earthquake damage scenarios on a reinforced concrete bridge in California.

Date: Wednesday, October 31, 2007
Time: 12-1 PM
Place: 721A Davis Hall
Title: Multiple support response spectrum analysis of bridges
Speaker: Katerina Konakli
Abstract: In this seminar, first, we briefly review the response spectrum method for multi-support seismic excitations and the extended method that accounts for the quasi-static contribution of truncated modes. We continue by presenting a parametric study of the correlation coefficients introduced in the extended method. The improvement by accounting for quasi-static contributions is examined for an example bridge structure. Finally, we introduce a coherency function that accounts for near-field effects and examine the structural response of an example bridge crossing the fault.

Date: Wednesday, October 24, 2007
Time: 12-1 PM
Place: 721A Davis Hall
Title: Design of reinforced concrete buildings by seismic envelopes
Speaker: Salvatore Sessa
Abstract: It will be shown a brief review about rectangular and supreme envelopes after a FEM modal analysis. A line search Newton-Raphson algorithm is used for evaluating the capacity domain edge and the forces' critical multiplier. The methods has been implemented in a software and it will be presented a complete seismic analysis of a strategic building following the EC8 code. It is possible to make a comparison between the results of the methods and qualitative consideration about the convergence behavior. Future work directions' hypothesis will be presented.

Date: Wednesday, October 3, 2007
Time: 12-1 PM
Place: 721A Davis Hall
Title: Random Vibration and Reliability Analysis in OpenSees
Speaker: Sanaz Rezaeian, Salvatore Sessa
Abstract: The purpose of this seminar is to present a brief orientation to the reliability module in OpenSees, and particularly focus on a newly developed sub-module called TELM (tail equivalent linearization method) that facilitates nonlinear random vibration analysis. To achieve this goal, first we will discuss the modeling process, choosing the excitation, and analysis possibilities in OpenSees. For random vibration analysis, OpenSees is used to obtain the necessary data (e.g. design points, TELSs,...), then matlab programs are used to perform linear random vibration analysis using the corresponding data. Several examples, including SDOF and MDOF systems excited by stationary and non-stationary processes, are presented to demonstrate the capabilities of this software.

Date: Wednesday, September 19, 2007
Time: 12-1 PM
Place: 721A Davis Hall
Title: Reliability Analysis and Design of Transmission Line Towers
Speaker: Jyothi Krishnan
Abstract: The seminar presents an assessment of the reliability index of tangent transmission line towers designed in accordance with current practice. Target safety factors, based on location of the tower and its importance, are recommended and effective wind load factors proposed to achieve these targets.

Date: Wednesday, September 12, 2007
Time: 12-1 PM
Place: 721A Davis Hall
Title: Multiple-support response-spectrum analysis of bridges
Speaker: Katerina Konakli
Abstract: This seminar will present a response spectrum method for seismic analysis of linear structures subjected to differential support motions, which is based on fundamental principles of random vibration theory. The original method is extended to account for the quasi-static contribution of truncated modes. A simple bridge model is considered to demonstrate the influence of spatial variability of the ground motion on selected response quantities.

Date: Wednesday, September 5, 2007
Time: 12-1 PM
Place: 721A Davis Hall
Title: Applicability of push-over analysis to plan irregular building structures
Speaker: Mario De Stefano
Abstract: As consequence of the increasing popularity of simplified non linear methods in seismic design, recently many procedures have been proposed aimed at extending push-over analysis to plan asymmetric structures. In particular, it appears very promising for practical purposes the approach based on combining pushover analysis of a 3D structural model with the results of a linear (modal) dynamic analysis. Nevertheless, a fundamental requirement for its effectiveness is that the elastic prediction of the envelope of lateral displacements be conservative with respect to the actual inelastic one. This reserch is aimed at verifying the above assumption with reference to both frame building structures and shear-wall ones. An extensive parametric analysis is conducted with simplified single-storey models by varying the main structural parameters that influence torsional response in the elastic and inelastic range of behaviour. It is found that the method is in general suitable for frame structures but it can not be recommended for shear-wall building structures, as their inelastic torsional response almost always exceeds the elastic one.

Date: Wednesday, August 29, 2007
Time: 12-1 PM
Place: 721A Davis Hall
Title: ICASP10 and Japan: a pictorial review
Speaker: Michelle Bensi and Sanaz Rezaeian
Abstract: The authors recently returned from Japan, where they attended the 10th International Conference on Applications of Statistics and Probability in Civil Engineering. In this seminar, they will make a review of their experiences in the conference and Japan by use of (a small fraction!) of their digital pictures.

Date: Thursday, July 26, 2007
Time: 11AM-12PM
Place: 721A Davis Hall
Title: Methods for reliability estimation and applications using FERUM: from FORM to MC-like simulation
Speaker: Jean-Marc Bourinet
Abstract: The presentation will provide an overview of a recent method developed at IFMA in France, improvements brought to existing methods and their implementations within FERUM. A snapshot of classical methods available within the FERUM version under development at IFMA will be given, including FORM and IS for multiple design points, subset simulation method, RBDO. A particular attention will also be paid to the calculation of FORM sensitivities, with respect to distribution parameters including correlation. A new method combining the subset simulation concept and a Support Vector Machines (SVM) learning method will also be detailed. This method addresses challenging problems, including system time invariant reliability problems and problems with a moderately large number of random variables (up to a few hundreds). Examples of applications will be presented, including recent FE based works on multiple crack propagation and buckling of imperfect shells. The presentation will also give a few details on FERUM capabilities for handling distributed computing on multi-processor computer platform and calling external FE codes.

Date: Monday, July 23, 2007
Time: 11AM-12PM
Place: 542 Davis Hall
Title: Application of Bayesian networks for seismic risk analysis
Speaker: Michelle Bensi & Daniel Straub
Abstract: The presentation shortly reviews a new framework for including structural reliability methods into Bayesian networks for modeling large and complex engineering systems. We will then focus on an example of seismic risk analysis of infrastructure systems, and in particular on the modeling of spatial and statistical dependences. The example will highlight the potential as well as the limitations of the new framework.

Date: Monday, July 16, 2007
Time: 11AM-12PM
Place: 721A Davis Hall
Title: New developments on a stochastic model for strong earthquake ground motion
Speaker: Sanaz Rezaeian
Abstract: Previously we discussed the development of a discrete and fully nonstationary model for strong earthquake ground motions. The model employs filtering of a discretized white-noise process and is fitted to target ground motions by matching a set of statistical characteristics. In this seminar we will look at the adequacy of ground motion models from a structural response point of view and will compare the response spectra computed from the simulated and target accelerograms.

Date: Monday July 9, 2007
Time: 11AM-12PM
Place: 721A Davis Hall
Title: Seismic critical response of polygonal reinforced-concrete sections
Speaker: Salvatore Sessa
Abstract: Seismic envelopes and iterative methods are combined to compute the ultimate strength of reinforced concrete sections. The results of preliminary testing are reported and investigation about the convergence of the algorithm is developed.

Date: Monday, July 2, 2007
Time: 1-2PM
Place: 721A Davis Hall
Title: Enhancing Bayesian Networks using Structural Reliability Methods
Speaker: Daniel Straub
Abstract: An overview on ongoing research on combining Bayesian networks with Structural Reliability Methods is presented. The results of some preliminary investigations are reported and the potential as well as the limitations of such an approach for risk analysis of civil systems are discussed.

Date: Thursday, May 24, 2007
Time: 1-2PM
Place: 721A Davis Hall
Title: Fire Risk Analysis by the Account Fire Model
Speaker: Kazuya Fujimura
Abstract: The Account Fire Model (AFM), which is being developed by Risk Management Solutions (RMS), will be discussed. The AFM provides the risk manager a view of risk that incorporates the probability of experiencing a range of losses from fire events, taking into consideration key characteristics that determine the extent of fire spread. The AFM uses a hybrid statistical and physical fire model designed to inform the primary insurance underwriting process, as well as facultative and treaty reinsurance management, and business continuity planning. With the AFM, Decision makers can be provided with risk metrics yielding an informed basis for pricing risk transfer, as well as evaluating the cost-benefit of capital investment to mitigate risk. Overview of physical fire modeling as well as statistical methods used in the AFM will be presented.

Date: Thursday, May 17, 2007
Time: 11AM-12PM
Place: 721A Davis Hall
Title: Bayesian methods in risk assessments for natural hazards
Speaker: Daniel Straub
Abstract: Bayesian methods are particularly useful in risk assessments when little information on the considered processes is available and when information changes with time, e.g., for weather-related events. These conditions are often encountered when modeling risks related to natural hazards. In this presentation, the fundamental principles of Bayesian methods and Bayesian networks will be reviewed and the possibilities of Bayesian networks as well as their limitations will be demonstrated by means of simple examples and illustrated by an overview on past and ongoing research and consulting projects that use Bayesian networks as a tool for natural hazard risk assessments.

Date: Thursday, May 10, 2007
Time: 1-2PM
Place: 721A Davis Hall
Title: Probability Density Evolution Method for Stochastic Mechanics
Speaker: Jianbing Chen
Abstract: Stochastic mechanics, which deals with randomness involved in excitations and structural properties of engineering systems, has gained increasing interest in the past decades. While a variety of valuable contributions have been made in some important aspects, the approach applicable to multi-degree-of-freedom nonlinear structural or mechanical systems has not been available as yet. The presentation will deal with a newly developed probability density evolution method. The principle of preservation of probability is revisited from the standpoint of state space description and random event description. Based on the random event description (Lagrangian description), a partial differential equation governing evolution of the one- or any arbitrary- dimensional joint probability density function can be established. Numerically solving this equation combining deterministic dynamic analysis with the finite difference method will yield instantaneous probability density function of any arbitrary response of interest. The probability density evolution method can be applied to static and dynamic stochastic response analysis of linear and nonlinear structures, dynamic reliability evaluation and system reliability evaluation, etc. Some key problems needing further investigations are discussed.

Date: Thursday, April 26, 2007
Time: 1-2PM
Place: 721A Davis Hall
Title: Airport Availability Modeling: A Different Perspective
Speaker: Jasenka Rakas
Abstract: This study proposes a systems-level approach to airport and runway availability assessments and prediction, and addresses the problem of the aging or continuously degrading aviation infrastructure.
Although the availability block diagrams are often used in the availability assessment of aerospace and electronic systems, their application to the airport availability problem on a system-level, developed in this study, is novel.� The proposed methodology is intended for short-term and long-term planning of the communication, navigation and surveillance (CNS) equipment acquisition, and investment and modernization decisions.
Improved understanding of the effects of equipment outages on airport availability is important in determining a required level of airport equipage and equipment reliability, particularly during critical operating conditions such as bad weather and increased traffic demand.� With proposed methodology an analyst can precisely quantify the additional level of airport availability achieved by upgrading or adding new pieces of CNS equipment.

Date: Thursday, April 12, 2007
Time: 1-2PM
Place: 721A Davis Hall
Title: Aleatory or epistemic? Does it matter?
Speaker: Armen Der Kiureghian
Abstract: The sources and characterization of uncertainties in engineering modeling for risk and reliability analyses are discussed. While many sources of uncertainty may exist, they are generally categorized as either aleatory or epistemic. Uncertainties are characterized as epistemic, if the modeler sees a possibility to reduce them by gathering more data or by refining models. Uncertainties are categorized as aleatory if the modeler does not foresee the possibility of reducing them. From a pragmatic standpoint, it is useful to categorize the uncertainties within a model, since it then becomes clear as to which uncertainties have the potential of being reduced. More importantly, epistemic uncertainties may introduce dependence between events, which may not be properly noted if their character is not correctly modeled. Influences of the two types of uncertainties in reliability assessment, codified design, performance-based engineering and risk-based decision-making are discussed. In particular, the influence of the dependence arising from epistemic uncertainties on systems and time-variant reliability problems are demonstrated through two examples.

Date: Thursday, April 5, 2007
Time: 1-2PM
Place: 721A Davis Hall
Title: Random sideway sliding of unlashed truck on a ro-ro vessel
Speaker: Peter Friis-Hansen, Technical University of Denmark
Abstract: Trucks on a ro-ro vessel are usually parked in parallel lanes very close to each other. To prevent movements of the truck wheels these are under severe weather conditions lashed to the deck. However, this lashing does not prevent that the top of the truck can be excited to vibrate, in particular in the side direction. If of some reason a truck is not lashed to the deck it may on top of its vibrations slide intermittently sideways in a random way. Thus this truck may bump into a neighbor truck and cause damage.
In the presentation the model will be described where the considered truck is idealized as a plane two degrees of freedom vibrating container carried by springs and dampers on a rigid undercarriage. The deck is assumed to move as a rigid body with two degrees of freedom, heave and roll, that together are modeled as a stationary Gaussian vector process. The problem is analyzed by use of crossing theory and the Slepian model process concept under the assumption that sideway sliding happens sufficiently rarely for the theory to be applicable.

Date: Thursday, March 22, 2007
Time: 1-2PM
Place: 721A Davis Hall
Title: Redundancy - je ne sais quoi?
Speaker: Charles Scawthorn, Kyoto University, Japan
Abstract: After a brief summary of activities at Kyoto University, the issue of lifeline and structural redundancy will be explored. Redundancy is of great value in achieving high reliability for networks and structures, yet is resistant to satisfactory quantification or even an adequate definition. Past research and definitions of lifeline and structural redundancy will be reviewed, followed by a summary of a recent (yet essentially inconclusive) research project.

Date: Thursday, March 15, 2007
Time: 1-2PM
Place: 721A Davis Hall
Title: Reliability-based acceptance criteria for deterioration limit states
Speaker: Daniel Straub
Abstract: The presentation will provide an overview on the issues involved in determining acceptability of deterioration in structural systems. Many owners of existing structures are confronted with the problem of determining whether their structure and their integrity management can be considered acceptable � however, to date, satisfying methods for demonstrating acceptability of the risks associated with deterioration in structural systems exist only for few special cases. Therefore, such methods have a huge potential and the aim of this talk is to present a proposal for such a simple yet consistent method to determine reliability-based acceptance criteria for deterioration in structural systems. First, it will be outlined what distinguishes deterioration limit states from ultimate limit states (i.e., failure of structural elements due to static or dynamic over-loading of the structure). In particular, the system and the uncertainty modeling required to account for deterioration is fundamentally different, as will be demonstrated by numerical examples investigating the influence of various factors related to deterioration. Thereafter, it will be demonstrated how reliability-based acceptance criteria, which were developed for ultimate limit states and which have been successfully implemented in several codes, can be adopted for deterioration limit states.

Date: Thursday, March 8, 2007
Time: 1-2PM
Place: 721A Davis Hall
Title: An introduction to support vector machines (SVM)
Speaker: Gang Li
Abstract: On the basis of a few available samples, to assign a new sample to classes (labeled numerically or symbolically) or to estimate the functional dependency of some variables with respect to others is the common problem in many areas. The former is the classification problem and the latter is the regression problem. Statistical learning theory (SLT) developed in the last decades is a powerful in solving these types of problems, in which support vector machines (SVM) is a popular tool since 1990s. This presentation introduces some related topics, including three parts: basic concepts of statistical learning, elements of support vector machines and discussion on reliability analysis.

Date: March 1, 2007
Time: 1-2PM
Place: 721A Davis Hall
Title: Probability-based vulnerability and criticality assessment of transportation infrastructure
Speaker: Michelle Bensi
Abstract: Recent events around the world have brought into light the need to understand and assess the vulnerability of civil infrastructure systems, specifically transportation infrastructure, to terrorist threats. While it has been accepted that large signature structures may be highly vulnerable as symbolic targets, an attack on a non-redundant, less visible link in an inventory such as a railway bridge or tunnel has the potential to cause substantial casualties as well as economic and social disruptions. It is not feasible to fortify and protect an entire inventory, and infrastructure owners must make decisions regarding how to best allocate their limited resources. This presentation will discuss the development of a probability-based method for quantifying the vulnerability of a public structure to attack and for ranking the criticality of a structure relative to the inventory by examining the consequences of an attack. Several illustrative case studies will be presented to demonstrate the methodology.

Date: February 8 and 15, 2007
Time: 1-2PM
Place: 721A Davis Hall
Title: Improved seismic fragility modeling from empirical data
Speaker: Daniel Straub
Abstract: The seminar presents the finalized version of the improved fragility model. The key feature of this model is that it accounts for statistical dependences among observations of seismic performances of system components, which have been neglected in past fragility analyses. As will be demonstrated by numerical examples considering seismic fragility of electrical substation equipment, the effect of this statistical dependence can be very large and must be considered when deriving fragility models from empirical data. The presentation concludes with an investigation of the effect of the model assumptions on the computed system fragility. It will be shown that the probability of system failure can change by orders of magnitude depending on the assumptions related to statistical dependences.

Date: February 1, 2007.
Title: New developments on a site-based fully nonstationary stochastic model for strong earthquake ground motion
Speaker: Sanaz Rezaeian
Abstract: In a previous seminar we discussed the development of a new discrete and fully nonstationary model for strong earthquake ground motions. The model employed filtering of a discretized white-noise process and was fitted to target ground motions by matching a set of statistical characteristics. In this seminar we will continue our discussion and will focus on the improvements that were made since last time. In particular, the topics of interest will be:
-Changes made to the form of the model and their advantages
-Modifications in parameter estimation when using statistical characteristics
-Ongoing developments on the weight functions in order to obtain a better fit of the actual recorded ground motion

Date: January 25, 2007.
Title: Dynamic optimal design for ice-resistant jacket platforms
Speaker: Gang Li
Abstract: The ice load is normally the dominant environmental force for offshore platforms in cold regions, which have to resist extreme static ice force and dynamic ice force. However, current design codes of ice-resistant structures only consider static ice forces. In recent years observations in Bohai Gulf showed that dynamic ice can induce significant vibrations, resulting in serious consequences. This presentation focuses on dynamic optimal design for ice-resistant jacket platforms, including the following two parts:
1) Minimum weight design of jacket platforms in consideration of fatigue life, in which a hybrid finite element model is adopted to simulate the jacket platform and the tubular joints, and the pseudo excitation method (PEM) is used to calculate the power spectral density of the hot spot stress.
2) Minimum dynamic response design of jacket platforms, in which a method is proposed to determine the most dangerous dynamic ice case in consideration of the randomness of ice thickness and velocity.

Date: December 6, 2006.
Title: Optimization of maintenance and replacement policies for a system of heterogeneous infrastructure facilities
Speaker: Samer Madanat
Abstract: This research addresses the determination of optimal maintenance and replacement policies for a system of heterogeneous facilities. The problem of optimizing maintenance and replacement policies at the system level is formulated in a reliability-based framework, based on policies that are optimal at the facility level. The facility-level policies are determined using a finite-state, finite horizon Markov decision process in which the state of the Markov chain contains information on the history of maintenance and deterioration.
Optimality conditions for the continuous-case system-level problem are derived and explained intuitively. A numerical study shows that the results obtained in the discrete-case implementation of the solution are valid approximations of the continuous-case results. The computational efficiency of the system-level solution makes the formulation suitable for systems of realistic sizes.

Date: November 29, 2006.
Title: Protecting Wetland Ecology by Using a Reliability-Optimization Framework to Manage Multiple Water Resources
Speaker: Chin Man Mok
Abstract: Tampa Bay Water is the largest wholesale water supply utility in Florida servicing the Northern Tampa Bay Tri-county area. To protect the wetlands and to prevent seawater intrusion while meeting the water demand of approximately two million people in the area, a reliability-based optimization framework was formulated for managing the available water resources from over 160 groundwater production wells in 12 inter-connected wellfields, 3 stream flow withdrawal facilities, a regional reservoir, and a desalination plant. The formulated framework computes the weekly groundwater production rates, surface water withdrawal rates, and reservoir water inflow/outflow rates that maximize the reliability of achieving healthy wetlands. The framework involves (1) a supply-system simulation model to represent the routing of available water resources, (2) Monte Carlo simulation models to generate conditional realizations of water demand and available surface water quantity based on a stochastic climatic model that accounts for the spatial and temporal characteristics, and (3) a stochastic unit response matrix (URM) model derived from a coupled HSPF surface water model and MODFLOW groundwater model that relates groundwater level response to groundwater extraction, and (4) a water-production operator simulator that mimic the response of the operators to meet the actual water demand in all circumstances. The reliability optimization problem is solved using a differential evolutionary (DE) algorithm.

Date: November 15, 2006.
Title: Seismic Risk Analysis of Buildings - source-focused formulation and business implementation
Speaker: Yuji Takahashi
Abstract: This seminar presents a source-focused formulation of seismic risk of buildings. In analyzing seismic risk, this equation enables to incorporate any renewal model, e.g. Poisson model, Brownian Passage Time model etc., for fault rupture occurrences. In addition, we can directly perform the end-to-end analysis from fault rupture to loss generation, and utilize simulation models developed in the related academic fields such as seismology, geotechnical engineering, and structural engineering.
In order to apply the proposed methodology to business, I launched and directed a cooperative research project that consists of twelve corporations. We developed commercial software for the risk analysis, and conducted case studies dealing with actual buildings. The case studies were performed based on two levels of the practical end-to-end analyses: time-history analysis and Capacity-Spectrum analysis. Some of the case studies are discussed in this presentation.
At the end of the presentation, I will briefly talk about my ongoing project entitled "Financial Technology to Promote Earthquake Preparedness in Buildings".

Date: October 18 and 25, November 1 and 8, 2006.
Title: A series of four lectures on: Introduction to Bayesian Networks for Engineering Risk Analysis
Speaker: Daniel Straub
Abstract: Bayesian networks (BN) are a highly efficient and intuitive tool to represent systems and processes subject to uncertainty. Because of their flexibility and computational efficiency, BN are increasingly applied in various areas of civil and environmental engineering for reliability/risk assessment and decision making. Examples of applications include: Reliability analysis of infrastructure systems, natural hazards risk assessment, probabilistic modeling in hydrology, optimization of construction operations, and risk based planning of inspection and maintenance. The seminar will provide an introduction to the use of Bayesian networks. Topics to be covered include:
  -  An introduction to reasoning under uncertainty and Bayes� theorem
  -  Modeling dependencies and identifying the network structure
  -  Decision optimization
  -  Computational aspects and limitations of the model
  -  Overview on available software for BN analysis
  -  Example applications
The content of the lectures may be adjusted according to specific interests.

Date: October 11, 2006.
Title: A stochastic model for earthquake ground motion with separable temporal and spectral nonstationarity
Speaker: Sanaz Rezaeian
Abstract: A discrete and fully nonstationary model for strong earthquake ground motions will be considered. In this model the discrete formulation and the ability of separating the temporal and spectral nonstationary characteristics allows flexibility and ease in modeling. The resulting nonstationary stochastic process can be fitted to a target ground motion in terms of a set of statistical data. The main purpose of this seminar is to focus on a discussion regarding the development of an optimization scheme that determines the parameters of the stochastic process using a set of target statistical characteristics which are obtained from previously recorded earthquakes in a region of interest.

Date: September 27, 2006.
Title: Hierarchical modelling and Bayesian inference of seismic fragility
Speaker: Daniel Straub
Abstract: Two different models for representing earthquake fragility from empirical data are presented and compared. One model is based on a limit state (engineering) approach, the second on a regression model (statistical approach). Both models are hierarchical, i.e. they differentiate between component-specific and site/system-specific factors, as opposed to traditional fragility models which neglect the system effects. Based on data of observed failures of equipement in electrical substations, it will be demonstrated how the appropriate model can be chosen. The Bayesian analysis of the data will be outlined and some of the difficulties involved will be discussed. The results obtained with different models will be compared. Finally, it will be demonstrated why the consideration of system effects (i.e. the use of hierarchical models) is crucial for system reliability assessments.

Date: August 16, 2006.
Title: Tail-equivalent linearization for nonlinear random vibration
Speaker: Kazuya Fujimura
Abstract: I will describe the main results of my doctoral dissertation.

Date: August 2, 2006.
Title: Effect of tuned mass damper on displacement demand of base-isolated structures
Speaker: Armen Der Kiureghian
Abstract: The effectiveness of a recently proposed scheme to reduce the displacement demand of a base-isolated structure by installing a tuned-mass damper underneath its isolation system is examined. Optimal parameters of the TMD for response to white noise are determined. It is found that, under such an excitation, a reduction of the order of 15% to 25% in the displacement demand can be achieved by adding the optimal TMD. Next, responses of an example base-isolated structure with and without an optimally designed TMD to selected suites of far- and near-field recorded accelerograms are compared. The study shows that for far-field ground motions the effectiveness of the TMD is more or less similar to that predicted by the white noise model, whereas for near-field ground motions the effectiveness of the TMD is far less, i.e., of the order of 10% or less. Reasons for this result are described.

Date: July 26, 2006.
Title: Uncertainty in failure rates of infrastructure systems
Speaker: Daniel Straub
Abstract: Reliability analysis of infrastructure systems (oil/gas, elctricity, transportation, water), often relies on failure rates which have been observed in the past. Unfortunately, the avilable data for estimating those failure rates is often sparse and the conditions under which these failures are observed are only partly known. For these reasons, the resulting failure rates are subject to large uncertainties. Based on the example of water distribution systems subject to earthquakes, a Bayesian model for representing the uncertainty on the failure rates will be presented, together with the tools for their evaluation. This is followed by a demonstration of the effect of the uncertainty on the overall system reliability and a discussion of how and when the presented models should be applied in practical risk management.

Date: July 6, 2006.
Title: A Nonstationary Stochastic Model for Strong Earthquake Ground Motion
Speaker: Sanaz Rezaeian
Abstract: In performance based earthquake engineering (PBEE), the use of actual accelerogram recordings have created questionable tasks about selecting and scaling recorded ground motions from different locations. Therefore, generating artificial ground acceleration time histories through a model that uniquely represents the characteristics of the subject site is of interest. This presentation focuses on developing a model that enables us to generate a target ground motion characterized by a finite number of parameters. The key to this method is the discrete format of the formulation that is achieved by filtering a discretized white-noise process. The advantage of this model over others is its ability to separately represent the temporal and spectral nonstationarity characters of the motion. This model can also be used for nonlinear random vibration analysis by the tail-equivalent linearization method.

Date: May 24, 2006.
Title: Some topics on life cycle cost evaluation
Speaker: Gang Li
Abstract: The evaluation of structural life-cycle cost is a key problem in performance based seismic design, which generally contains three major problems: modeling the evaluation formulation, calculating the occurrence probability of failure modes of interest, and determining the loss value of the failure mode when it occurs. This presentation focuses on the first two problems with the following topics:
-- expected loss evaluation in consideration of correlation between failure modes
-- expected loss evaluation based on fuzzy theory
-- approximate efficient reliability analysis of some specific types of structures

SPECIAL EVENT: Reliability Symposium: Presentations by students in CE229. See program.
Date: May 19, 2006
Time: 9:45AM to 3:40PM
Place: 544 Davis Hall

Date: May 10, 2006.
Title: An overview on reliability and risk based inspection planning (RBI)
Speaker: Daniel Straub
Abstract: In RBI, Bayesian decision analysis is applied to optimize the inspection efforts on deteriorating structures, by explicitly addressing the uncertainties both in the deterioration processes and the inspection performance. The presentation will give a concise overview on RBI, including
    -   The general decision optimization problem
    -   The probabilistic modeling of the deterioration and the inspections
    -   Computational issues (structural reliability analysis vs simulation methods)
    -   Risk acceptance
    -   System aspects
An overview on practical applications of the methodology to the planning of inspections on offshore structures subject to fatigue concludes the talk.

Date: April 21, 2005.
Title: Structural-reliability-based analysis of large and/or complex systems
Speaker: Charles Menun
Abstract: The probability that a large and/or complex system of interconnected facilities (e.g., power and water distribution networks and transportation systems) fails to perform in a satisfactory manner during and after an extreme event such as an earthquake or hurricane is naturally of interest. Structural-reliability-based procedures are well suited for such analyses because, in addition to providing an estimate of the probability that a system will fail, they also provide sensitivity information that can be used to identify those components within the system that are most critical to its performance. However, there are a number of algorithmic problems that must be resolved before available system-reliability-based analysis procedures can be efficiently applied to large and/or complex systems of components. For example, if the inclusion-exclusion rule is used to compute the probability of failure of a series system, then the required number of probability calculations increases geometrically with the number of components present in the system, thereby limiting the applicability of such an approach. This paper describes the required modifications to conventional structural-reliability-based calculations for the performance assessment of large and/or complex systems of components. Numerical examples are provided to illustrate the application of the proposed procedure.

Date: April 14, 2005.
Title: PDF of maximum response by the TELM
Speaker: Kazuya Fujimura
Abstract: In the PBEE application, the PDF of maximum response (displacement, velocity, acceleration) is required for a given level of intensity measure. The simulation approach using actual earthquake records selected by the results of PSHA can be used for this purpose. However, this approach includes the subjectivity due to the selection of earthquake records, which might result in a large variability in the final output of the PBEE application. To avoid this subjectivity, a random vibration approach is proposed as an alternative. Under stationary excitation, the PDF can be obtained efficiently by using TELM. An example application to a 6dof shear-building model will be presented.

Date: April 7, 2005.
Title: Reliability Methods in Medical Decision Making
Speaker: Lilit Yeghiazarian, Department of Bioengineering, UCLA
Abstract: Most medical decisions are normally made without the benefit of complete information, the treatment entails incomplete efficacy along with costs and risks, and consequently the assurance of successful cure is never perfect. The uncertainty is associated not only with the validity of the diagnostic tools, but also with the inherent randomness and variability in the neuro-physiological organization of individual patients, and the probabilistic nature of physical and biological parameters associated with tumor development and invasion. We propose to extend reliability methods, in particular FORM/SORM, to cancer analysis and risk assessment to enhance medical decision making. Applications in multiscale cancer models are also considered.

Date: March 17, 2005.
Title: SORM and Simulation in the Tail Equivalent Linearization Method (TELM)
Speaker: Kazuya Fujimura
Abstract: The accuracy of the TELM depends on the accuracy of the FORM for the instantaneous failure probability. When the limit-state surface is far from being flat, FORM may not provide a good approximation, and consequently TELM may not provide a good approximation to the first passage probability. One possible solution is to apply SORM or Simulation to achieve a better approximation to the instantaneous failure probability. An example problem will be shown to discuss the property of the solution.

Date: March 3, 2005.
Title: Identification of Critical Components and Cut sets Using Linear Programming Bounds on System Reliability
Speaker: Junho Song

Date: February 24, 2005.
Title: Tail-equivalent linearization for nonlinear random vibration
Speaker: Kazuya Fujimura

Date: February 17, 2005.
Title: The inclusion-exclusion formula and its use in estimating system reliability
Speaker: Junho Song

Date: February 10, 2005.
Title: Risk analysis using Bayesian network
Speaker: Armen Der Kiureghian
Abstract: A brief introduction to the use of Bayesian networks (BN) for risk analysis will be presented. The Hugin program developed in Denmark for BN analysis will be demonstrated.