Development of an Open-Source Autonomous Computational Fluid Dynamics Meta-Modeling Environment for Small-Scale Combustor Optimization

Abstract

Abstract This work presents an open-source autonomous computational fluid dynamics (CFD) metamodeling environment (OpenACME) for small-scale combustor design optimization in a deterministic and continuous design space. OpenACME couples several object-oriented programing open-source codes for conjugate-heat transfer, steady-state, multiphase incompressible Reynolds averaged Navier-Stokes CFD-assisted engineering design metamodeling. There are fifteen design variables. Nonparametric rank regression (NPRR), global sensitivity analyses (GSA), and single-objective (SOO) optimization strategies are evaluated. The Euclidean distance (single-objective criterion) between a design point and the utopic point is based on the multi-objective criteria: combustion efficiency (η) maximization and pattern factor (PF), critical liner area factor (Acritical ), and total pressure loss (TPL) minimization. The SOO approach conducts Latin hypercube sampling (LHS) for reacting flow CFD for subsequent local constraint optimization by linear interpolation. The local optimization successfully improves the initial design condition. The SOO approach is useful for guiding the design and development of future gas turbine combustors. NPRR and GSA indicate that there are no leading-order design variables controlling η, pattern factor (PF), Acritical , and TPL. Therefore, interactions between design variables control these output metrics because the output design space is inherently nonsmooth and nonlinear. In summary, OpenACME is developed and demonstrated to be a viable tool for combustor design metamodeling and optimization studies.