Adjoint-Assisted Topology-Optimization-Inspired Analysis of Pseudo-Porous Flow Fields: Application to a Flettner Rotor

TL;DR

This paper introduces an adjoint-assisted, topology-optimization-inspired method for analyzing flow sensitivities in porous media, providing a non-intrusive, qualitative tool for early-stage flow improvement assessment in fluid dynamics applications.

Contribution

The paper presents a novel, non-intrusive approach to evaluate topological sensitivities in fluid flows using adjoint methods without modifying existing solvers, aiding flow optimization insights.

Findings

Method accurately predicts flow sensitivity regions.

Validated on a Flettner rotor in atmospheric wind.

Supports early-stage flow design improvements.

Abstract

This paper presents an adjoint-assisted, topology-optimization-inspired approach for analyzing topological sensitivities in fluid domains based on porous media formulations -- without directly utilizing the porosity field as a design variable. Instead, the method evaluates the sensitivity with respect to the porosity parameter via an indirect, adjoint formulation, enabling intuitive visualization of cost-functional improving or deteriorating regions. These sensitivity fields are not used in a classical optimization loop but rather to qualitatively assess spatial areas that indicate potential for flow improvement. A key benefit of the proposed approach is its non-intrusive nature: no modifications to existing adjoint solvers, e.g., for shape optimization purposes, are required. As such, the method can be interpreted as an alternative post-processing step that supports early-stage design insights. The approach is demonstrated on a Flettner rotor operating in atmospheric wind at a diameter-based Reynolds number and spinning ratio of ReD=2E+06 as well as k=3 to highlight its practical relevance and interpretability. Placing a container stack in cost-functional improving or deteriorating areas validates the method's prediction quality of topological flow sensitivities.