Advances in analytical and computational processes improve vehicle optimization
By Giuseppe Cataldo|December 2024
The Multidisciplinary Design Optimization Technical Committee provides a forum for those active in development, application and teaching of a formal design methodology based on the integration of disciplinary analyses and sensitivity analyses, optimization and artificial intelligence.
In February, researchers at ONERA, the French national aerospace research center, released the Surrogate Modeling Toolbox 2.0. This new version includes features that have proven highly useful for engineering design in general and for the multidisciplinary design optimization community in particular, especially for modeling, optimization and uncertainty quantification purposes.
In June, the Multiscale Multiphysics Design Optimization Laboratory at the University of California, San Diego released ParaLeSTO-COMSOL, an extension of the Parallel Level-Set Topology Optimization code introduced last year. ParaLeSTO-COMSOL leverages COMSOL’s graphical user interface to streamline the setup of finite element analysis, while a short MATLAB script integrates this process into an optimization framework. Designed to foster multidisciplinary collaboration and enhance education in topology optimization, the tool has been integrated into UC San Diego’s graduate-level courses on topology optimization and additive manufacturing. It has proven effective for tackling complex optimization challenges, including plasticity and conjugate heat transfer with turbulence.
Researchers at MIT’s Engineering Systems Laboratory in Massachusetts continued to develop miniMDO Python with funding from Airbus. This software package facilitates the formulation of monolithic MDO architectures for problems with analytic disciplines, objectives and constraints. The package also has a module to reformulate the problems based on sparsity structure. The codebase, which is still experimental, can be found on GitHub.
Researchers in the Computational Optimal Design of Engineering Systems Laboratory at the University of Arizona developed a multifidelity approach for the construction of surrogate aerodynamic databases based on computational fluid dynamics simulations and wind tunnel tests. Through an adaptive sampling and model fidelity management scheme, the Gaussian process-based approach limits the number of expensive simulations or tests while accounting for various sources of uncertainty. This year, researchers applied the approach to heat flux predictions on high-speed vehicles.
In July at NASA’s Goddard Space Flight Center in Maryland, a group of researchers, in collaboration with the NASA-funded Jet Propulsion Laboratory in California and Bocconi University in Italy, upgraded the Quantification of Uncertainty Analysis Toolkit, or QUAnT. This set of computational tools efficiently quantifies uncertainty to guide the design process of complex, multidisciplinary systems throughout their lifecycles and under limited resources. Based on multifidelity modeling and efficient sampling techniques, QUAnT performed global sensitivity analyses and risk assessments for NASA’s Mars Sample Return program, demonstrating improvements in computational cost and maximizing the value of information that can be extracted from a limited number of tests due to budget and schedule constraints.
In January, researchers at Virginia Tech, in collaboration with Northrop Grumman, developed an advanced framework for multidisciplinary structural analysis and design optimization of spacecraft structures by innovatively integrating commercial software and open-source Python libraries to facilitate easy implementation in aerospace industries. The framework can efficently optimize composite space vehicle structures for a range of mechanical, thermal and acoustic loads, with the help of an innovative approach that uses lamination parameters and mixed integer programming techniques to reduce optimization costs by some 60% compared to widely used meta-heuristic algorithms.
In September, Virginia Tech researchers also demonstrated an efficient parametric reduced-order modeling technique to optimize composite structures. They achieved a 97% reduction in analysis cost and a 99.9% reduction in memory requirements compared to its full-order counterpart in optimizing tow-steered fiber composite panels. The researchers are also developing an MDO code for aircraft design and performance analysis using liquified natural gas. LNG is environmentally friendly and cost-effective, has 26% more energy density than classical jet fuel and has vast reserves compared to currently used aviation fuels.
Contributors: Youssef Diouane, Rakesh Kapania, Alicia Kim, Samy Missoum and Johannes Norheim
