Research
Many flows of fundamental interest and engineering applications are complex due to complicated geometry and/or the presence of a range of length and time scales, which is a well-known characteristic of turbulent flows. Accurate simulation of such flow problems requires high-fidelity numerical methods like direct numerical simulation (DNS) and large eddy simulation (LES), which traditionally have been restricted to fairly simple geometries. My research focuses on understanding complex physics of turbulent flows relevant to engineering applications using the state-of-the-art numerical simulations on some of the largest clusters in the country.
Despite being one of the most studied canonical fluid problems, turbulent boundary layers (TBL) are far from being fully-understood, mainly in the presence of pressure gradient and transverse curvature which are commonly encountered in many engineering applications. My research attempts to combine boundary layer theory with the latest simulation techniques to enhance our current understanding of such flows.
Many problems of practical interest involve turbulent flows at high Reynolds numbers, which are not feasible for DNS. LES of such flows can be computationally feasible if the energy-containing scales are resolved and the effect of small scales is modeled using what is called subgrid model. There are numerous subgrid models proposed in literature, each with its own merits and demerits. However, none of them exhibit all the desirable properties, such as numerical stability and high correlation with the true subgrid stress. Moreover, the energy-containing scales of wall-bounded turbulent flows become increasingly smaller near wall as Re increases, making even LES challenging and impractical in such cases. This problem can be alleviated if the requirement of resolving energy-containing near wall scales is relaxed, as done in wall-modeled LES where the effect of unresolved near-wall small scales is modeled using what is called a wall model. The accuracy of wall-modeled LES rely heavily on the accuracy of the underlying wall model. There are numerous wall models proposed in literature, but most of them assume equilibrium and neglect pressure gradient effects, and are thereby limited to simple flow problems. Wall-modeled LES of complex flows demands a wall model which is local in nature, with minimal assumptions and simple implementation with small computational overhead for a general unstructured LES flow solver. My research focuses on evaluating the existing subgrid and wall models in terms of their performance in complex flows and developing future models to overcome the deficiencies of the current popular models.