In the field of thermal protection, detailed three-dimensional computational
fluid dynamics (3D-CFD) simulations are widely used to analyze the thermal
behavior on a full vehicle level. One target is to identify potential violations
of component temperature limits at an early stage of the development process. In
battery electric vehicles (BEVs), transient load cases play an increasing role
in evaluating components and vehicle systems close to real-world vehicle
operation. The state-of-the-art 3D simulation methodologies require significant
time and computational effort when running transient load scenarios. One main
reason is the conjugate characteristic of the problem, meaning that conduction
within the component and convection into the surrounding air occur
simultaneously. This requires a detailed consideration of both the fluid and
structural domains.
Therefore, this article derives a time-efficient simulation methodology for
transient component temperatures in electric vehicles. The approach is to
extract heat transfer coefficients and reference temperatures from sample flow
simulations and to construct convective meta-models. Solid component
temperatures are then transiently computed whereby the low-dimensional
meta-models provide the convective heat transfer. Dimensional analysis
determines the smallest possible parameter space for the meta-modeling. Two
different types of meta-models, a scalar regression model and a vector proper
orthogonal decomposition (POD) approach, are tested and compared.
The study examines at first the applicability of the heat transfer formulation
under different flow and component temperature conditions using a generic flat
plate test case. A low Biot number (Bi) is crucial to receive accurate
temperature predictions as heat transfer coefficients are derived on uniform
temperature walls. The methodology is subsequently applied to a sample component
in the motor compartment. Measurements on a test rig and a transient load case
comparison with a coupled simulation prove the validity of the numerical
procedure. Scaling to full-vehicle applications is feasible. The new methodology
delivers a highly accurate temperature prediction and increases computation
efficiency, especially for sensitivity studies.
