TITLE:

LS-OPT©: Methods and Applications in MDO and Material Identification.

DATE:

Friday, March 22nd, 2013

TIME:

3:30 PM

LOCATION:

GMCS 214

SPEAKER:

Nielen Stander.
Livermore Software Technology Corporation Livermore, California, USA.

ABSTRACT:

LS-OPT© is the parametric multi-disciplinary optimization (MDO) package which accompanies the
multi-physics code LS-DYNA©.

The most common applications of LS-OPT are in design optimization, reliability and material
calibration. To tackle these problems, LS-OPT provides a choice of optimization strategies. Two broad
methodologies are available namely Direct and Metamodel-based Optimization.

Metamodel-based Optimization is commonly executed using the Sequential Response Surface
Method (SRSM) in which an approximate model is updated using iterations of carefully selected
multiple simulation sets. A metamodel optimization is conducted at each iteration as a predictor for
the next cycle. Although other metamodels are available, the most common surrogates are Radial
Basis Function Networks and polynomials. Multi-Objective Optimization can be conducted using either
Metamodel-based or Direct Optimization via the NSGA-II algorithm.

After a detailed review of the optimization methodology, an example involving six load cases with
highly detailed automotive crashworthiness and NVH models is discussed.

A second design example is presented to demonstrate the Reliability-based Design Optimization
(RBDO) of a vehicle by incorporating variation of the input parameters as well as a specification of the
probability of failure.

Finally, a material identification example is presented to show how physical experimental results can
be used to calibrate complex material models with hysteretic behavior.

In conclusion an overview of current and future developments is given.

HOST:

Dr. Satchi Venkataraman.

DOWNLOAD: