Quantifying parametric uncertainty in complex engineering systems is critical in large-scale simulations as it provides confidence levels and moreover it can probe the sensitivities of the system and the nonlinear interactions of its components. There has been great progress on uncertainty quantification (UQ) in the last ten years, especially at the component level, but there several open questions remaining for complex systems characterized by many random parameters (high-dimensionality) for forward but also backward UQ, i.e. parameter estimation.  Advanced stochastic approximation techniques are necessary to tame the computational cost associated with high-dimensionality and inverse problems so that numerical solutions become feasible at reasonable computational cost. This mini-symposium will explore recent advances in numerical algorithms and applications for high-dimensional uncertainty quantification, model reduction, and stochastic inversion in large-scale high-dimensional complex systems.

This minisymposium is devoted to UQ of coupled problems, including multiscale, multiphysics, multidomain, and discrete-continuum problems, as well as to UQ in structural dynamics.  The various aspects of UQ are of interest, including stochastic modeling (excitations, parameters, modeling errors, noisy measurements), numerical methods for uncertainty propagation through high-dimensional and coupled problems, and statistical inverse methods and data assimilation.  All application areas are of relevance, such as structures, materials, fluid-structure interaction, vibroacoustics, biomechanics.

The performance of engineering structures subjected to dynamic loads are subjected to a variety of uncertainties arising from material properties, topology, loading, physical modelling mechanisms, etc. Uncertainties can have significant economical implication on the design of modern dynamic-prone structures often featured by creative topology, light-weight materials but met with high performance standards. Quantifying and managing them is becoming increasingly important in modern engineering projects. This mini-symposium aims at providing a forum for knowledge exchange of the latest developments in the quantification and management of uncertainties related to structural dynamic applications, spanning over aspects of theoretical, computational, experimental and practical nature. Both forward prediction and backward identification problems are covered. Topics relevant to the minisymposium include, but not limited to:

• Modelling of uncertainties related to structural dynamics
• Stochastic dynamics
• Reliability analysis of structures subjected to dynamic loads
• Modal identification
• Structural system identification and damage detection
• Vibration testing, sensor and actuator configuration design

Papers dealing with experimental investigation and practical applications are especially welcome.

This is a proposal for the organization of a minisymposium at the ECCO-MAS International Conference on Uncertainty Quantification in Computational Sciences and Engineering (UNCECOMP) on 25-27 May 2015, Crete Island, Greece. The aim of this minisymposium is to compare different Uncertainty Quantification (UQ) methods and approaches that are being used and developed for computational sciences. These UQ methodologies include, but are not limited to, probabilistic frameworks, interval approaches, fuzzy techniques, and combinations of those. Within these frameworks, Galerkin formulations will be compared with advanced sparse grid collocation methods and other response surface methods for the propagation of uncertainties. Contributions in any of these topics are welcome in this minisymposium and, in particular, contributions that compare multiple techniques. The objective of this minisymposium is to determine the suitability of different UQ techniques for solving various problems in the computational sciences and to learn from each other's approaches towards uncertainty propagation. Applications to uncertainty reduction and management, optimization under uncertainties, and verification and validation in any branch of scientific computing are encouraged, with an emphasis on computational fluid and structural dynamics, and heat transfer. The minisymposium will consists of at least six presentations on these topics.

In computational science and engineering, numerical models of physical systems are constructed to reproduce experimental observations and other empirical evidence. The parameters of the numerical models are determined by combining information from different sources such as direct measurements of the parameters or the system behavior, expert knowledge, categorical data and information from literature. In probability theory, the process of combining information to learn model parameters is formalized in the concept of Bayesian updating. Thereby, the prior probability distribution of the model parameters is updated with new data to a posterior distribution. The derived distribution can be further used for forward uncertainty propagation and reliability assessment of the system performance.

This mini-symposium aims to attract papers that address either methodological developments or novel applications on Bayesian analysis of numerical models. Individual relevant topics include: Markov chain Monte Carlo methods; sequential Montel Carlo methods; Taylor series approximations to the posterior; Kriging/Gaussian process models; conjugate priors and Gibbs sampling; approximate Bayesian computation; structural identification; reliability updating; updating in the presence of spatial/time variability; updating of meta-models; applications that investigate the influence of prior considerations on the analysis results; definition of the likelihood function; representation of model errors

Many materials of importance in engineering and science are heterogeneous on a range of scales, thus making it difficult to treat them via homogenization approaches. The appearance of inelastic dissipative behavior (plasticity, damage, cracks) additionally introduces a strong dependence across all scales due to localization. Furthermore, the structure at small scales is typically described in terms of statistical or stochastic models, introducing uncertainties which – due to effects like localization – may not average out on macroscopic scales, but may require a stochastic model on microscopic scales as well. There are currently many different methods to transfer the description and physical behavior between different scales under development, as well as numerical methods being explored to implement these transfers efficiently. We welcome contributions on modelling elastic and inelastic or energy conserving and dissipative behavior of engineering and geotechnical materials by taking into account the heterogeneities of real materials. We especially welcome the contributions on different reduction methods that are trying to connect different scales of material observations and produce new models capable of accounting for the uncertainties in the material response.

Real engineering structures are designed to fulfill prescribed safety requirements under environmental loads, such as earthquake ground motion, sea waves, gusty winds etc., which have an intrinsic random nature. In the last decades, several studies have been devoted to the development of efficient procedures for the evaluation of the failure probability of structures subjected to stochastic excitations. The most widely used failure model identifies failure with the first exceedance of a specified threshold. Unfortunately, analytical methods can be applied only in very special cases. Approximate methods and simulation techniques are needed to perform reliability assessment of real engineering systems.

Structural reliability analysis under both stochastic excitation and parameter fluctuations due to inherent uncertainties in material and geometrical properties has also attracted much research interest. Furthermore, extensive studies have been devoted to reliability sensitivity analysis and reliability-based optimization procedures.

The aim of this Mini-Symposium is to present the latest advances in the field of structural reliability analysis under environmental loads modeled as stochastic processes. Contributors are invited to focus on theoretical and computational issues related to stochastic analysis and design in the context of reliability assessment of real engineering problems.

Topics may include, but are not limited to:

• Random Vibration (linear and nonlinear)
• Stochastic Processes
• Stochastic differential calculus
• Wind Engineering
• Earthquake Engineering
• Offshore and Marine Structures
• Reliability analysis under uncertainties
• Reliability sensitivity analysis
• Reliability-based optimization

n recent years the interest and research activity in uncertainty quantification (UQ) for complex systems modeled by partial differential equations (PDE) has increased significantly. This motivation has roots in our lack of knowledge about model parameters, source terms, coefficients, initial/boundary conditions for most practical applications, such as in computational fluid dynamics. Such a trend calls for the incorporation of stochastic components in predictive simulations and optimizations - in order to properly model uncertainties in complex systems - and for the adoption of stochastic methodologies able to cope with computationally intensive models. Progress was made in part due to growing available computing resources as well as new efficient numerical methods for high-dimensional problems. Assuming all input uncertainties are correctly identified, they can be propagated forward through simulations in order to quantify their effects. However UQ remains computationally costly for most realistic industrial applications. One way to reduce the uncertainty and dimensionality is to assimilate additional – potentially noisy, indirect and incomplete – data to the model in the form of observations, experiments, expert judgment… This is a so-called inverse problem of identification that is often ill-posed and requires regularization procedures. Data assimilation techniques can be considered as the probabilistic formulation of an inverse problem and are therefore broadly categorized into methods based on estimation theory, control theory, and stochastic approaches. The Bayesian framework is for instance a central tool in estimation theory and is at the core of Bayesian inference and Bayesian filtering. However, these techniques remain costly and the computational effort associated with the numerical solution of an inverse problem is typically larger by at least an order of magnitude than that of the corresponding forward problem. The development of robust and accurate probabilistic reduced-order models and reduced stochastic models to approximate the high-dimensional state of a dynamic system with a low-dimensional approximation in a subspace of the state space is therefore crucial.

For instance, in a Bayesian inference framework, one may identify three major computational tasks that may benefit from reduced-order / surrogate modeling: 1. Estimation of the a priori probability density of the input parameters based on all the prior information; 2. Computation of the likelihood function that describes the interrelation between the observation and the unknown; 3. Developing methods to explore the posterior probability density of the quantity of interest, especially in the context of high-dimensional problems. Typical techniques to accelerate statistical inference rely on computationally efficient surrogate models such as stochastic spectral methods, Gaussian process model or projection-based reduced-order models. A legitimate question relates to the level of fidelity of the approximate model in order to retain an accurate posterior distribution; in particular when the forward model exhibits strong nonlinear behavior with respect to its inputs.

The objective of this mini-symposium is to bring together experts in computational techniques for forward and inverse problems in a stochastic framework in order to discuss recent advances in the field. We encourage the participation of researchers interested in a broad range of applications, such as data assimilation, stochastic optimization, robust design/optimization, geosciences, computational fluid dynamics, oceanography, civil engineering systems, climate, material science, combustion...

In recent years, the tremendous advances in technology have provided us with sensors that are smaller, low-cost, and energy efficient. Beyond the MEMS revolution, a plethora of sensing solutions has emerged based on wireless technology, fiber-optics, principles of optics and acoustics, nanotechnology, bio-inspired sensors and more. By coupling ease of deployment, cost efficiency, smart capabilities and sufficient accuracy such solutions have opened the way for reducing uncertainty of engineering systems though local or global, non-destructive monitoring schemes.

This mini-symposium will form a platform for idea exchange and knowledge dissemination concerning the latest developments in the field of innovative sensor solutions and their implementation on structural or mechanical systems. Topics relevant to the minisymposium include, but are not limited to, implementations and algorithmic solutions for:

• Wireless sensors
• Fiber optical sensors
• Ultrasonic/ Acoustic emission sensors
• Self Sensing sensors
• Bio inspired sensors
• Non contact sensors

Contributions pertaining to actual implementation of such sensor solutions on the non-destructive evaluation and characterization of large scale systems are especially welcomed.

Uncertainty inherently exists in measuring, testing and modeling of physical systems across multiple length and/or time scales. Over the last few years, integration of stochastic methods into a multiscale framework or development of multiscale models in a stochastic setting for uncertainty quantification, analysis and design of physical systems is becoming an emerging research frontier. Continuous efforts in this direction are deemed essential to bridge the gap between current "ad hoc" phenomenological models and the multiscale nature of various complex problems of mechanics, e.g. fatigue, dynamical fracture/fragmentation or earthquake prediction. The objective of this Mini-Symposium is to reflect the recent research efforts and progress towards mingling of multiscale analysis and design with uncertainty quantification, and to bring together researchers seeking interactions among multiscale methods and computational mechanics in order to obtain reliable predictions of the behavior of physical systems.

Non probabilistic  approaches  like interval  and  fuzzy  techniques  are becoming  increasingly  popular  in  the  context  of  modelling  for  engineering applications.  These methods are capable  of  handling  limited  data,  and  are  often  considered to  be  very  appropriate  in  an  engineering  context  when  a  full  probabilistic  quantification  of  the  model  uncertainty  is  not  available,  or  alternatively,  when  a  full  probabilistic  quantification  of  the  analysis  result  is  not  required. This  mini-symposium focuses  on  the  application  of  these  techniques  for  the  representation  of  uncertainty  in  typical  engineering  modelling  activities.  Researchers focusing  on  the  application  of  interval  and  fuzzy  techniques in  numerical  modelling  for  engineering  applications,  ranging  from  uncertainty  propagation  methodologies,  identification  and  quantification techniques to  optimization  under  uncertainty  are  invited  to  submit  an  abstract  to  this  mini- symposium.