Real-Time Aero-Icing Simulations Using Reduced Order Models

3D fully viscous turbulent aero-icing simulation remains too computationally intensive when broad range parametric studies are needed, such as during a certification process or anti-icing systems optimization. In order to make such simulations more practical, this work presents a reduced order modeling based on the Proper Orthogonal Decomposition (POD) method which is a popular approach to extract a low-cost CFD approximation from a limited number of "snapshots". POD, equipped with multidimensional Kriging interpolation method, is able to predict the uncalculated solution at intermediate values of three and more input parameters. In this work, this approach is applied to calculate in-flight icing. Also presented are the examples of other current and future applications of POD in aero-icing including droplet impingement calculation and Computational Wind Engineering. The method of selecting the pre-computed CFD solutions (snapshots) and the performance of the Kriging interpolation method are identified as the two main factors which affect the quality of POD approximations. The smoothness of POD coefficients alpha assumed during Kriging interpolation is examined.