Parametric model order reduction for large-scale and complex thermal systems
For engineers simulating complex thermal systems, this provides a computationally efficient reduced model that preserves parametric dependence, though the approach is incremental.
The paper proposes a parametric model order reduction technique for large-scale thermal systems, enabling direct updates of physical parameters in the reduced model. The method achieves moment matching with a derived error bound.
In this paper, a parametric model order reduction (pMOR) technique is proposed to find a simplified system representation of a large-scale and complex thermal system. The main principle behind this technique is that any change of the physical parameters in the high-fidelity model can be updated directly in the simplified model. For deriving the parametric reduced model, a Krylov subspace method is employed which yields the relevant subspaces of the projected state. With the help of the projection operator, first moments of the low-rank model are set identical to the correspondent moments of the original model. Additionally, a prior upper bound of the error induced by the approximation is derived.