Scope: This minisymposium convenes leaders of several concurrent development efforts in uncertainty quantification (UQ) methodologies, techniques, and tools for design and manufacturing process engineering. The emphasis is on practical methodologies and techniques that are effectively implemented by mathematically rigorous, computationally efficient, and engineer-friendly tools. The solutions should advance systems engineering and/or concurrent engineering approaches. The target audience is engineering practitioners who would like to better solve their UQ problems, as well as researchers who would like to understand the scope of, and relationship between, existing state-of-the-art UQ solutions in design and manufacturing process engineering.
Specific issues to address include:
- Uncertainty analysis (propagation of uncertainties) and sensitivity analysis in (potentially complex) systems of systems, esp. for "black-box" deterministic function models, but also for stochastic models such as discrete event simulation.
- Tolerancing, stochastic optimization, reliability, risk management, and robust decision making (e.g., robust product performance and ease/cost of manufacture).
- Modeling: inverse problems/model calibration (e.g., Bayesian parameter estimation), use of surrogate models, validation, and integration with design of experiments.
- Treatment of aleatory/irreducible, epistemic/reducible, and other uncertainties such as model and computational uncertainties.
- Managing/handling missing data and information (e.g., incompletely specified distributions).
- Managing/handling dependencies in multidimensional inputs and outputs.
- Analytical vs. numerical techniques, including mathematical rigor (theoretical underpinnings and implementation assumptions) and computational cost.
- Maturity of, and availability of an open interface with, the supporting software toolset, and its ease of use.
Session 1:
- Exact calculations for random variables in uncertainty quantification using a computer algebra system, by Larry Leemis (College of William and Mary). The Maple-based APPL language performs exact probability calculations involving random variables. This talk introduces the langauge and presents several applications, including bootstrapping, goodness-of-fit, reliability, probability distribution selection, stochastic activity networks, time series analysis, and transient queueing analysis.
- HPD Liberates Applied Probability AND Enables Comprehensively Tackling Design Engineering, by Jean M. Parks and Chun Li (Variability Associates). HPD, which stands for Holistic Probabilistic Design, consists of a methodology and two breakthrough software suites for probabilistically tackling problems of any complexity in Design Engineering. Initially developed at Xerox Corporation where variability of thousands of characteristics affecting system performance dominated product development efforts, the HPD software, in fact, liberates Applied Probability and has exposed shortfalls of existing techniques! As importantly, HPD revolutionizes Stochastic Analysis and Optimization for manufacturing industries!
- Practical UQ for Engineering Applications with DAKOTA, by Brian M. Adams, with Laura P. Swiler and Michael S. Eldred (Sandia National Laboratories). Through DAKOTA, we strive to deploy general-purpose UQ software to a broad range of engineering applications, supporting risk-informed design. Practical engineering UQ methods must manage simulation cost, nonlinearity, non-smoothness, and potentially large parameter spaces. Emerging reliability, stochastic expansion, and interval estimation algorithms in DAKOTA address these challenges. These can be employed in mixed deterministic/probabilistic analyses such as optimization under uncertainty. This talk will highlight application and environment complexities that can limit efficient uncertainty quantification.
- Open TURNS: Open source Treatment of Uncertainty, Risk 'N Statistics, by Anne-Laure Popelin, Anne Dutfoy, and Paul Lemaitre (EDF R&D, Industrial Risk Management Department). The needs to assess robust performances for complex systems have lead to the emergence of a new industrial simulation challenge: to take into account uncertainties when dealing with complex numerical simulation frameworks. Open TURNS is an Open Source software platform dedicated to uncertainty propagation by probabilistic methods, jointly developed since 2005 by EDF R&D, EADS Innovation Works, and PhiMECA. This talk gives an overview of the main features of the software from a user's viewpoint.
- UQ practices and standards for design and manufacturing process engineering, by Mark Campanelli (National Institute of Standards and Technology). Design and manufacturing process engineering have always necessitated some degree of uncertainty management. Increasingly complicated systems, with greater performance requirements/constraints, require corresponding advances in uncertainty quantification (UQ) practices. To make effective engineering decisions, these practices are typically implemented with software tools that must address multiple types of uncertainty across a spectrum of models, measurements, and simulations. This talk discusses the present evolution of UQ practices and the potential for improvement through standards development.
- Systematic Integration of Multiple Uncertainty Sources in Design Process, by Sankaran Mahadevan (Vanderbilt University). A design optimization methodology is presented that systematically accounts for various sources of uncertainty – physical variability (aleatory uncertainty), data uncertainty (epistemic) due to sparse or imprecise data, and model uncertainty (epistemic) due to modeling errors/approximations. A Bayes network approach effectively integrates both aleatory and epistemic uncertainties, and fuses multiple formats of information from models, experiments, expert opinion, and model error estimates. The methodology is illustrated using a three-dimensional wing design problem involving coupled multi-disciplinary simulation.
- Generalized Chapman-Kolmogorov Equation for Multiscale System Analysis, by Yan Wang (Georgia Institute of Technology). In multiscale system analysis, the effect of incertitude due to lack of knowledge may become significant such that it needs to be quantified separately from inherent randomness. A generalized Chapman-Kolmogorov equation based on generalized interval probability is proposed to describe drift-diffusion and reaction processes under aleatory and epistemic uncertainties. Numerical algorithms are developed to solve the generalized Fokker-Planck equation and interval master equation, where the respective evolutions of the two uncertainty components are distinguished.
- A Comparison Of Uncertainty Quantification Methods Under Practical Industry Requirements, by Liping Wang, Gulshan Singh, and Arun Subramaniyan (GE Global Research, USA). The objective of this paper is to establish comprehensive guidelines for engineers to effectively perform probabilistic design studies through a systematically comparison of uncertainty quantification or probabilistic methods. These methods include 1) Simulation-based approaches such as Monte Carlo simulation, importance sampling and adaptive sampling, 2) Local expansion-based methods such as Taylor series method or perturbation method, 3) First Order Second Moment (FOSM) based methods, 4) Stochastic expansion-based methods, such as Neumann expansion and polynomial chaos expansion, 5) Numerical integration-based or moments-based methods, such as Point Estimate Method, Eigenvector Dimension Reduction, 6) Regression- or metamodel-based methods, 7) Bayesian methods, and 8) Data classification methods. Both aleatory & epistemic uncertainty will be addressed in the paper. A set of benchmark problems representing different levels of dimensionality, non-linearity and noise level will be used for the comparison.
Links to conference website and submission info.
Organizer:
Mark Campanelli
NRC Postdoctoral Fellow – Process Engineering Group
Systems Integration Division (SID), Engineering Laboratory (EL)
The National Institute of Standards and Technology (NIST)
100 Bureau Drive, Stop 8263
Gaithersburg, MD, 20899-8263, USA
Email: mark.campanelli@nist.gov
Phone: (301) 975-4461
