Abstract
Turbulence remains to be one of the most complex phenomena observed, particularly in the industry. Representation and resolution of turbulent fluid flow continues to be difficult. This can be attributed to the nature of turbulence which is characterized by constant occurrences of irregularities and fluctuations of fluid flow parameters with respect to space and time. Despite the tremendous progress in the field of Computational Fluid Dynamics (CFD) that has given rise to various models and theories that attempt to explain turbulence, including frequently used simulation methods like the Direct Numerical Simulations (DNS) and the Large Eddy Simulations (LES), these methods fall short due to various shortcomings in each individual method. With the advent of Big Data and rapid advances in the field of Machine Learning, it has now become feasible to develop frameworks of learning models that can be custom tailored to suit the substantial amount of data required to solve turbulence and also address the spatiotemporal nature of the data. The rise in the usage of data driven models and learning algorithms have provided a platform for new and improved ways in which turbulence can be understood and resolved. The aim of this research is to use advanced neural networks, a machine learning approach, in order to enhance and facilitate the computation of turbulent fluid flows so that it would pave the way in the development of a system that would yield results at acceptable tolerance range, optimal resolutions at low computational cost and time required.