Page 5 - ASME_SMASIS_2018_Program

Special Events

SHORT COURSE

Course Title: Applications of Uncertainty Analysis in Smart Materials and Adaptive Structures

Sunday, September 9th
1:00 PM - 4:30 PM
Presidio A

Description
The purpose of this hands-on tutorial is to expose participants to statistical and numerical techniques that will allow them to quantify the accuracy of
multi-physics models and simulation codes for active materials and structures when one accounts for uncertainty or errors in models, parameters,
numerical simulation codes, and experimental data.

In the first part of the tutorial, we will provide an overview of Bayesian statistics and numerical algorithms necessary to propagate input uncertainties
through simulation codes. We will consider several case studies to illustrate these techniques for a variety of materials and smart structure applications.
These include models for piezoelectric macro-fiber composites, shape memory alloys, viscoelastic polymers, and quantum-informed ferroelectric
continuum models. In this part of the tutorial, we will provide participants with algorithms that quantify the uncertainties in model parameters, such as
piezoelectric constants, when they are calibrated using experimental data. To illustrate the uncertainty propagation techniques, we will demonstrate
the construction of 95% prediction intervals for PZT models at a given applied field. Secondly, we will demonstrate the use of sensitivity analysis to
determine which model parameters are influential in the sense that they can be quantified from data and which are noninfluential and should be fixed at
nominal values during uncertainty quantification.

As part of the tutorial, we will have participants run case studies using MATLAB. These studies will include models and data provided by the instructors,
but participants are also encouraged to bring their own models and data for testing during the tutorial, based on their specific problems of interest.

Learning Outcomes

• Utilize sensitivity analysis and Bayesian statistics to identify and quantify the uncertainty of model parameters using experimental data or higher-fidelity
model predictions.

• Implement techniques to propagate uncertainties associated with model parameters and experimental observations through models to compute
distributions or prediction intervals for responses or quantities of interest.

Intended Audience

The intended audience is for graduate students, industrial practitioners, and academic professionals who are interested in quantifying uncertainty in
material and structural models using experimental data or higher-fidelity model predictions.

Course Level
This course will be given at the Introductory level. Basic knowledge of probability and MATLAB programming is required.

Course Length
Half day (3.5 hours); 0.35 CEU

Biography

This course will be co-taught by Professors Ralph C. Smith and William Oates. Dr. Smith is a Distinguished Professor of Mathematics at North Carolina
State University, who has expertise in mathematical modeling, uncertainty and sensitivity analysis, and control of smart materials and structures. He
has written books on both smart materials and structures as well as uncertainty quantification and sensitivity analysis. He has investigated the role of
uncertainty quantification in the context of macro-fiber composites and shape memory alloys including the use of uncertainty quantification to improve
robust control design. Dr. Oates is a Professor in the Department of Mechanical Engineering at Florida State University. His research includes constitutive
model development, structural analysis, and experimental characterization of smart materials and adaptive structures. He has utilized Bayesian statistics
to analyze smart materials and systems including quantum informed ferroelectric modeling, graphene thermoacoustics, piezoelectric composites, and
multi-functional polymer constitutive model development.

Additional Comments

It is recommended, but not required, to have a laptop with a MATLAB license to take full advantage of the hands-on portion of the tutorial. Also
recommended, but not required, is experimental data that has already been fit to a model; e.g., via optimization.

5