A state observer based data assimilation method between RANS and robotic PIV data

Abstract

Experimental fluid dynamics and computational fluid dynamics have traditionally been treated as disparate fields of study. However, each field has its own unique set of advantages and disadvantages. Data assimilation is a field that can be used to leverage some of the advantages each field offers to help compensate mutual
weaknesses. In this thesis, a state observer based data assimilation method is used to assimilate 3-D experimental data obtained in a wind tunnel experiment onto a steady RANS simulation. The experimental data is considered as the ground truth and is used to condition the RANS simulation. An understanding of the working of the method along with a study on the effect of different parameters of the state observer method are gathered by first applying it on the 1-D viscous Burgers equation and a 2-D CFD simulation. For the 3-D case, experimental data is obtained by performing a wind tunnel experiment using robotic PIV to map the time-averaged velocity field around a bluff body following which the data is assimilated onto a steady RANS simulation of the same body. Application of this method helps to recreate topological features and velocity
fields of the flow with better accuracy than a baseline CFD simulation. Finally, the effects of the different parameters on the success of the method along with recommendations for improving the method are provided.