Modeling of Flashback With Different Blends of CH4 and H2 by Using Finite Rate Chemistry With Large Eddy Simulation

Abstract

Due to its clean combustion characteristics, hydrogen fuel is gaining attention in power generation. New designs of engine systems and components are being explored to allow blending with the increasing amount of hydrogen in natural gas. Adding H2 increases the probability of flashback and often is one of the main constraints in using high H2 blends in premixed combustors. There are several mechanisms of flashback like boundary layer flashback, combustion induced vortex break down, turbulence in the flow, fluctuations in equivalence ratio, etc. Semi-empirical models, based on non-dimensional numbers and flame speed, have successfully predicted flashback propensity for a given operating condition. The semi-empirical models are computationally very efficient; however, they lack generality. A typical combustor can have multiple flashback mechanisms. The relative importance of each mechanism can change with a change in the combustor design or even with a difference in the operating conditions for the same combustor. Since prediction of flashback requires accurate modeling of highly transient chemistry phenomena and the impact of heat loss on chemistry, a viable detailed chemistry solution is preferred to model flashback. This paper describes the use of a finite rate chemistry model to predict flashbacks in a turbulent premixed combustor in this work. The configuration used is a swirl stabilized combustor (SimVal) from National Energy Technology Laboratory. The current computations are done with Finite Rate Chemistry (FRC) and Large Eddy Simulations (LES). Simulations are carried out for a varied percentage of CH4/H2 blends, ranging from 0% H2 to 100% H2 at a fixed equivalence ratio and inlet mass flow. As the percentage of H2 is increased in the fuel, flame speed also increases. With this, the propensity for flashbacks also increases. A 28-species reduced mechanism for CH4/H2 blend flames is used to keep the simulations computationally tractable. The simulations with the reduced mechanism are performed by considering non-adiabatic effects from heat loss near the walls and multi-component property considerations. This improves the accuracy of the FRC-LES simulations to capture the onset of boundary layer flashback towards the inlet. The simulations from FRC-LES suggest a fine mesh in the boundary layer for an accurate prediction that makes the simulations expensive. Therefore, an Adaptive Mesh Refinement (AMR) approach has been used for different CH4/H2 blends to accurately model the flashback without any loss in generality as the AMR criteria used here are applicable for a wide range of conditions. The FRC-based solution strategy proposed in this work provides a framework to model flashback for different blends without any case-specific tuning.