CONTROL VARIATE POLYNOMIAL CHAOS: OPTIMAL FUSION OF SAMPLING AND SURROGATES FOR MULTIFIDELITY UNCERTAINTY QUANTIFICATION

We present a multifidelity uncertainty quantification numerical method that leverages the benefits of both sampling and surrogate modeling, while mitigating their downsides, for enabling rapid computation in complex dynamical systems such as automotive propulsion systems. In particular, proposed utilizes intrusive generalized polynomial chaos to quickly generate additional information is highly correlated with original nonlinear system. then leverage Monte Carlo-based control variate correct bias caused by approximation. contrast related works merging adaptive approximation setting, (gPC) selected because it avoids statistical errors design providing analytical estimates output statistics. Moreover, enables theoretical contributions provide an estimator strategy optimally balances computational efforts allocated gPC construction. deploy our approach multiple examples including simulations hybrid-electric systems, where shown achieve orders-of-magnitude reduction mean squared error statistics estimation under comparable costs purely or approaches.