CFC2023

CFD Model Order Reduction for Optimizing Ice Protection Systems

  • Pourbagian, Mahdi (McGill University)
  • Habashi, Wagdi (McGill University)

Please login to view abstract download link

Due to its multiphysics nature, the numerical simulation of in-flight icing requires the three-step successive calculation of airflow, droplet impingement, and conjugate heat transfer to yield the accreted ice shape. Admittedly, the majority of CFD industrial icing analyses are still based on 2D calculations over airfoil sections. 3D approaches are assumed to require substantial computational resources and solution times for each step of the three-step calculation, with the combined resources for a truly 3D three-step calculation considered prohibitive. The challenge is even more formidable when dealing with optimizing ice protection systems, a procedure requiring thousands of parametric calculations thus placing it beyond today’s computing capabilities. Reduced order modeling (ROM) is shown to be the path that ought to facilitate the adoption of 3D calculations for in-flight icing. ROM uses a limited number of snapshots (truly 3D CFD-Aero or CFD-Icing solutions obtained at various operating conditions) to extract, via proper orthogonal decomposition (POD), a basis of modes that represent its most fundamental physical features. A linear combination of these modes, with coefficients obtained via multi-dimensional interpolation, can then be used to rapidly (minutes, and sometimes seconds) obtain fully 3D solutions for all conditions within the design space other than the initial snapshots. This paper presents a short description of the POD-based ROM procedure and then focuses on its applications to the optimization of aircraft and rotorcraft anti-icing and de-icing hot-air and electrothermal systems. The applications demonstrate that ROM is the facilitator of supposedly prohibitive numerical simulations. When the snapshots are handled by CFD, a ROM-based “CFD simulator” can be delivered to non-CFD experts to be used without CAD resources, painfully long mesh generation, solver monitoring, and parameter adjustments. Such an approach could be an enabler to a multitude of second-tier suppliers who may not have the human, software, or hardware capabilities to carry out such intricate analyses, allowing them to match in geometry and quality the calculations ran by the OEM on the same aircraft.