Uncertainty Quantification of MEMS Devices with High-Dimensional Correlated Process Variations

This work reports a new technique to solve high-dimensional uncertainty quantification (UQ) problems for MEMS device design. By introducing GLD-Spearman transformation and compressed sensing (CS) to polynomial chaos expansion (PCE), this reported method achieves UQ for high-dimensional non-Gaussian correlated variants without rebuilding the basis polynomial functions Therefore, our method is highly efficient. It is verified by the analysis and experiment of the MEMS test structure for thermal expansion coefficient.