Automated Geometric Optimization of Turbulent Airflow in Ducts Using Genetic Algorithms

Many industrial, automotive and aerospace systems generate power from fluid flows. A common challenge in designing such components is the routing of flow within a duct or pipe geometry. However, routing is often considered secondary to the design of main components and therefore subject to compromises. Further, optimal routing designs are difficult to achieve within these constraints using current engineering tools, and are often a manual process involving computer aided design (CAD) and computer aided engineering (CAE) packages. An automated tool is therefore required for the design and optimization of flow ducts and pipes in both the conceptual design and engineering phases. In this work, a fully automated tool chain is developed for the design and optimization of fluid routing based on minimal pressure drop. The tool, built in C++, utilizes the ACIS geometry modeling engine and interfaces with Gambit for unstructured grid generation. A steady-state flow analysis is performed using incompressible turbulent flow via the Fluent computational fluid dynamics (CFD) package. Optimization is controlled by the GAlib library of genetic algorithms (GA) which have been adapted to interface with the numerical description of the geometry. This paper outlines the process undertaken by the tool chain to reach an optimal pipe design, given the boundary sizes and positions in space, as well as constraints on the system. Specific case studies are presented for smooth pipe turbulent airflow. The case studies focus on using splines to define and parameterize the geometry. The resulting optimal duct geometries are verified with structured and grid independence analysis methods. Finally, the paper discusses wider applications of the developed tool chain to more complex routing fluid flow problems, and non-routing based applications.