Comparative Evaluation of Flamelet Generated Manifold and Finite Rate Combustion Model Results for Bluff Body Stabilized Pure Hydrogen Flame Using NTNU Atmospheric Rig Test Data

Abstract Hydrogen combustion is taking center stage due to the benefits of low greenhouse gas emissions and high energy density but comes with its own set of challenges (viz. auto-ignition, flashback, safety and flame stability). Accurate numerical modelling of hydrogen combustion is hence essential, and this paper focuses on comparing the results between the flamelet generated manifold (FGM) model and finite-rate (FR) combustion model for the conical bluff body stabilized premixed hydrogen flame operating at a lean condition (0.4 equivalence ratio) using the NTNU Trondheim workshop data (based on the atmospheric test rig installed at the Norwegian University of Science and Technology). This paper highlights the results obtained from Finite Rate (FR) and FGM setup, along with the incremental changes to FGM model by including (a) non-unity Lewis number flamelets (b) Turbulent flame speed closure model (TFSC) against the Finite rate closure model for Progress variable source term (c) Strained flame speed model which generates a laminar flame speed lookup table as a function of strain (d) Indirect inclusion of preferential/differential diffusion using user defined function (UDF) for hydrogen flames. Each of the above FGM edits has different influences on flame shape, heat release and velocity prediction. Finite rate model captured the V flame shape and showed a good match with experimental data as against the baseline FGM prediction of M flame shape; the incremental FGM edits helped to bridge the deficiency and predicted comparable results to FR and experimental data. The impact of the unity Lewis number approximation with the baseline FGM model was clearly visible in the equivalence ratio and temperature distribution near the flame front region. Ansys Fluent’s Native GPU solver for the FR combustion model is also explored in this paper; with the intent to compare the speedup between CPU and GPU solver.