Streamlined structural design optimization with interactive visualization for multi-objective design spaces

Abstract

A common task for design engineers is the balancing of conflicting design objectives through the application of design of experiment (DOE) based methods to explore a design space. Traditional approaches to this problem require modeling the geometry with graphical user interface (GUI) based modeling software to generate the DOE runs. Visualizing these design tradeoffs, particularly when there are more than two objectives, is also a challenging task that often needs to be done manually. This thesis provides a streamlined approach to structural design optimization with interactive plotting for multi-objective design spaces, enabling engineers to easily visualize the objective space and identify optimal solutions. Utilizing Python and its available libraries, in conjunction with the innovative Jupyter Notebook environment, parameterized geometry is generated using Python code, and a static structural analysis is performed through a DOE. The resulting dataset from the DOE features five output objectives and is used in an optimization process generating Pareto optimal solutions, and thus a Pareto front for given pairs of constraints is generated. An interactive 2D scatter plot is generated with the Pareto front data, allowing users to quickly investigate the relationships between the inputs and output objectives. Additionally, two approaches for visualizing the tradeoff between more than two objectives are presented, enabling an informed decision to be made when selecting a candidate solution when balancing more than two objectives. These approaches are demonstrated on two example structural design problems. Finally, the potential application of large language models (LLMs) to the code based approach used in this thesis is discussed.