Development of an Open-Source Autonomous CFD Meta-Modeling Environment for Small-Scale Combustor Optimization – Part II

Abstract

This work presents an improved open-source autonomous CFD meta-modeling environment (OpenACME) for small-scale combustor design optimization. OpenACME couples several object-oriented programming open-source codes for CFD-assisted engineering design meta-modeling. OpenACME is fully automated and can be used in PC workstations or HPC clusters. OpenACME uses a global metaheuristic optimization algorithm based on multiple-objective evolutionary algorithm (i.e., NSGA-II). An initial design population is first computed with Latin Hypercube Sampling and subsequent iterations generate offspring based on tournament mating, uniform crossover, and polynomial mutation. OpenACME is capable of computing multiple parallel CFD design points concurrently, speeding up the meta-simulations. The CFD are based on steady-state, incompressible, three-dimensional simulations with multi-phase k-ω SST RANS and “frozen” flamelet progress variable (FFPV) combustion model. There are fifteen design variables. There are three meta-simulations. The meta-simulations report Pareto Frontier from which optimum designs can be selected based on thermodynamic cycle requirements. Conjugate heat transfer provides the most realistic liner temperature and combustor performance. Acritical is still recommended as a cost function when liner durability is a concern. OpenACME demonstrated to be a viable tool for combustor design optimization.