Random-space dimensionality reduction scheme for expedient analysis of microwave structures with manufacturing variability

A dimensionality reduction scheme is presented for the expedient statistical analysis of microwave structures exhibiting manufacturing variability in geometric and electrical parameters. In the proposed approach, the computational complexity of the high-dimensional random space that is often necessary to describe the stochastic electromagnetic boundary-value problem is mitigated by employing a principal component analysis with sensitivity assessment in combination with an adaptive sparse grid colocation scheme. The proposed method exploits the inherent dependencies between parameters to reduce the number of simulations needed to extract the statistics of desired output response parameters. The attributes of the method are demonstrated through the analysis of the cross talk between the wires of a coupled stripline transmission line structure.