Sparse Grid Surrogate Models for Electromagnetic Problems With Many Parameters

A sparse grid surrogate model (or metamodel) is proposed to reduce the computational time involved by accurate electromagnetic (EM) simulators. Sparse grids have already been used in many applications for interpolation and integration. The method can treat a high ber of independent parameters that are intractable for many other techniques due to the curse of dimensionality. Beyond the conventional method, adaptive sparse girds are also studied. The capabilities are illustrated through the examples drawn from the EM nondestructive evaluation.