High Dimensional Uncertainty Quantification in Integrated Micro-System Multiphysics Modeling

In real-world scenarios, the true state of the integrated micro-system is influenced by various uncertainties, rendering it infeasible to assess micro-system performance with deterministic multiphysics modeling alone. Therefore, we propose a high-dimensional uncertainty quantification framework, which describes the multiphysics fields of integrated microsystems from a statistical perspective. Considering the impracticality of intrusive methods due to multiphysics coupling and geometric uncertainties, the proposed method relies solely on data, ensuring its applicability. To address the curse of dimensionality caused by high-dimensional uncertainties, we evaluate the correlation between polynomials and the fields of interest, retaining only significant bases to reduce the required sample size. Numerical experiments demonstrate the superior performance of the proposed algorithm in extracting statistical features, predicting multiphysics field, and analyzing sensitivity.