A Scalable Conjugate Heat Transfer Method Applied to Engine Transients

Abstract This paper introduces a scalable implementation of a loosely coupled approach for solving unsteady conjugate heat transfer (CHT) problems. The framework uses segregated solvers; computational fluid dynamics (CFD) to solve the Reynolds-averaged Navier-Stokes (RANS) equations on the fluid side, and an iterative finite element analysis (FEA) method to solve the heat equation on the solid side. Thermal coupling is achieved through an iterative procedure that applies temperature (Dirichlet) boundary conditions on the fluid side, and convective (Robin) boundary conditions on the solid side. It is assumed that the disparity in characteristic timescales between fluid and solid is such that the fluid can be treated as quasi-steady, permitting the use of steady CFD. The loosely coupled approach allows simulations to time-march at time steps suitable for the solid side, offering at least an order of magnitude speed-up compared to a fully coupled approach, which is limited by the stability of the fluid side. For validation purposes, the framework is applied to two test cases: a flat plate and a rotating cavity with radial inflow. The convergence characteristics for a range of relaxation levels are compared. Under certain conditions, as has been reported in the literature, loosely coupled CHT methods can become unstable. We discuss these instability mechanisms, and offer practical guidelines for maintaining stability for complex cases. Computational scalability is demonstrated using a 100+ million cell full-annulus radial inflow cavity model, which is run through an engine representative transient in ∼ 24 hours.