Subject: ✈ MDOlab Newsletter

View this email online if it doesn't display correctly
MDOlab Newsletter—Fall 2014
MDOlab group photo with Boeing 777
Dear Friend
Welcome to the MDOlab newsletter, an update on research and open source software that we send a few times a year. You are receiving this because I think you are interested in numerical optimization, MDO, engineering design, or aircraft design. If this is not the case, feel free to unsubscribe. If you know someone who might like to subscribe, please forward them this newsletter. Best regards,
Latest publications
Wing aerodynamic shape optimization benchmark
RANS-optimized wing
The AIAA Aerodynamic Design Optimization Discussion Group developed a series of benchmark cases. In this paper, we solve the RANS-based wing optimization problem, try to find multiple local minima, and solve a number of related wing design optimization problems. The initial and optimized geometries and meshes are provided here.

Aerodynamic design optimization of a blended-wing body aircraft
BWB 3D-printed models
This builds on our previous work on stability-constrained flying wing optimization. A series of RANS-based aerodynamic design optimization studies shows the tradeoffs between drag, trim, and stability for the NASA/Boeing BWB. The photo on the left shows 3D-printed models with pressure colormaps.

Satellite multidisciplinary design optimization benchmark
Optimized CRM
In collaboration with NASA and the Michigan Exploration Lab, we developed a new large-scale benchmark MDO problem, and solved a problem with 25,000 design variables and 2.2 million state variables by optimizing the data downloaded from a CubeSat subject to operational and physical constraints. This problem is now a plugin in the OpenMDAO open source project. 

Parallel structural finite-element analysis for design optimization
TACS wingbox model
The Toolkit for Analysis of Composite Structures (TACS) was developed by Prof. Graeme Kennedy, and is the structural solver used in the MDOlab. Its main advantages are the parallel scalability and efficient gradient computation via a parallel adjoint method.

Open-source software
pyOpt: A common Python interface for nonlinear optimizers 
Benchmark wing design optimization problem
This is a Python wrapper for a number of optimizers that makes it easy to setup optimization problems and switch between optimization algorithms. pyOpt enabled us to do the comparison of 8 optimizers for a wing optimization problem, including gradient-based and gradient-free algorithms.

GeoMACH: A geometry engine for aircraft configurations
GoeMACH aircraft configurations
This is a geometry engine that enables the rapid generation of parametric aircraft outer mold lines and internal structures, which facilitates high-fidelity MDO. GeoMACH computes all the geometry derivatives, so it is well suited for use with gradient-based optimizers.

OpenMDAO: A framework for multidisciplinary optimization
XDSM diagram
This framework is being developed by NASA to facilitate the implementation of MDO problems. It also provides a number of built-in algorithms for the solution of such problems. The latest version provides an efficient method for computing coupled derivatives for gradient-based optimization.
Other recent publications
A survey of MDO architectures
MDO architectures
There are a number of architectures available to solve multidisciplinary design optimization problems, but so far, they had not been explained in a consistent notation. This survey paper describes all known MDO architectures and provides flow diagrams for each architecture.

High-fidelity aerostructural optimization of aircraft configurations
Aerostructural model
We have enabled design optimization of lifting surfaces with respect to hundreds of aerodynamic and structural design variables. The numerical methods used to achieve this are described in the first paper, and the second one applies these methods to the multipoint optimization of a transport aircraft.

A review of methods for computing derivatives
Optimized CRM
This paper reviews all available methods for computing derivatives numerically (finite-differences, complex-step, adjoint method, and algorithmic differentiation), and unifies these methods using a single equation.

You may unsubscribe or change your contact details at any time.