ISBN 978-3-8439-5262-0

978-3-8439-5262-0, Reihe Fahrzeugtechnik

Yannis Werner
An Approach for the Handling of Geometric Constraints in Engineering Shape Optimization

194 Seiten, Dissertation Technische Universität Braunschweig (2022), Softcover, A5

Zusammenfassung / Abstract

Computer-aided development methods are becoming increasingly important in today’s product development process. Engineers commonly use CAD and FEM-based systems for structural design and dimensioning. Today, the design is manually adapted to the calculation results iteratively until they meet all stiffness, failure, and package requirements. Often, multi-criteria requirements for the structure must be fulfilled, e.g., stiffness and crash requirements in the car body area. These constraints commonly contradict the geometrical constraints of the package. Today’s optimization programs usually do not offer any form of structural optimization, enabling multidisciplinary optimization under the accountancy of geometrical constraints. Implementing such an optimization routine provides considerable potential for cost and time savings in designing structural components. Such structural models can be optimized using modern metaheuristic optimizers and optimization algorithms. Such a program is implemented here with Python and thus can optimize structural models quickly. The author derives a method for detecting external geometries using ANSA. Different models are set up, and the method is validated for different "real-life" scenarios, like a b-pillar side impact and a longitudinal crash-member test. Both approaches are implemented in the framework. The method is approved and tested in different scenarios. Initial tests validate the NSGA-II algorithm for the method, as it considers the geometries and improves the structural performance. As crash calculations are very time-consuming, different methods for reducing the number of numerical calculations for crash load cases are tested. The comparison identifies the GOMORS framework as a highly potential algorithm for the application. The implementation of the framework shows that surrogate-assisted algorithms can significantly improve the optimization speed, even under the feasibility constraint.