Dynamic multidisciplinary design analysis and optimization workflows: Concept and implementation

• Dynamic workflows allow changing design variables, tools and constraints at runtime • Dynamic MDAO workflows enable system architecture optimization in a single loop • New methodology to formulate, visualize, exchange and materialize dynamic workflows • Switches, branches, and subworkflows enable dynamic behavior within workflows • Less runtime and fewer function evaluations for architecture design optimizations Many architectural design trade-offs must be performed during the conceptual design of complex systems, such as aircraft, to identify promising design concepts. A few architectures are usually selected and traded to keep the process manageable, but this can suffer from biases. Ideally, an optimizer, supported by a multidisciplinary system evaluator, should enable architecture design space exploration. However, incorporating architectural design choices into a multidisciplinary design optimization is challenging due to changing design variables, constraints, and disciplinary tools associated with different architectural designs. This paper proposes a new formal methodology for dynamic Multidisciplinary Design Analysis and Optimization (MDAO) workflows. These workflows allow the design variables, tools, and constraints to change during execution, enabling evaluation and optimization of different architectures within a single MDAO system definition. Switches, branches, and subworkflows are introduced to enable dynamic behavior within the workflow. Together, they allow for a complete mathematical problem definition and consistent formalization. Accordingly, extensions to the eXtended Design Structure Matrix (XDSM) and the Common MDO Workflow Schema (CMDOWS) are presented to enable the visualization, storage, and exchange of dynamic workflows. An automated MDAO formulation, integration, and execution process is extended to ease the setup of these workflows. The methodology has been verified and validated using a mathematical optimization problem from literature. The dynamic workflow successfully discovered a Pareto front comprising multiple architectural design options. Furthermore, the results demonstrate that a single dynamic workflow can identify optimal architectures more efficiently than multiple static workflows, requiring significantly less execution time and fewer function evaluations.