Statistical Emulator Construction for Nonlinear Smart Systems

Comprehensive physical models can accurately quantify the dynamics of nonlinear and hysteretic systems but often require significant computational cost. This can reduce their effectiveness when performing sensitivity analysis, uncertainty analysis, parameter calibration or system design which typically requires multiple iterations of computationally expensive routines. This can also preclude the use of these models for real-time model-based control design. Emulators provide statistical approximations to comprehensive physical models which provide two advantages: high efficiency and statistical characterization of missing model components. We discuss the construction of statistical emulators to provide efficient surrogates for nonlinear smart material models. We will primarily focus on emulators for the homogenized energy model for ferroic compounds.