This project aims to improve our fundamental understanding of turbulent premixed combustion. We consider the basic problem of a statistically planar flame front propagating on a homogenous/isotropic turbulence in a periodic rectangular domain. The current project is part of an on-going series of direct numerical simulations(DNS) studies with a step-by-step complicated flame models, starting with the simplest model of flame being an infinitely-thin interface propagating with a fixed speed relative to the flow and with a constant density and viscosity across the flame front, then allowing the flame front to have finite-thickness, then allowing density variation and non-unity Lewis number diffusion, and finally enabling complex fuel chemistry. The computational tasks in this project consist two parts: the first part is to perform DNS caculations for new cases, the second part is data analysis on the obtained DNS database. For data analysis, the machine learning (ML) approach (based on the Google Tensorflow software) will be applied, which will be benefited from the GPU nodes in Tegner cluster.