Simao Marques

In aerospace design, achieving the optimum outcome often involves multiple iterative design rounds, necessitating a balance between objectives and constraints. However, traditional aerodynamic design optimization, particularly in transonic regime, is hindered by its high computational cost. To meet the evolving needs of aircraft design, we present an efficient alternative approach for performing high-fidelity aerodynamic design optimization. The discrete empirical interpolation method (DEIM) is used to obtain a set of interpolation points. These points are utilized to reconstruct flow fields using bases obtained from a database compiled from high-fidelity analyses. The aerodynamic reduced-order model (ROM) is then built based on a Global Local neural network, which introduces a nonlinear mapping function between reduced design variables and the DEIM interpolation points. This data-based ROM offers cost savings compared to the iterative use of Computational Fluid Dynamics (CFD) simulations typically employed in gradient-based design optimisation. The sensitivity analysis is performed on each component that contributes to the development of the ROM. The ROM is utilized to solve the transonic optimisation problem for the undeflected Common Research Model (uCRM). The results obtained using ROM are similar in shape to those obtained using full-order model-based optimisation, with a difference in CD of about two counts. These findings demonstrate the effectiveness of the proposed method for fast and high-fidelity aerodynamic design.

In aerospace design, achieving the optimum outcome often involves multiple iterative design rounds, necessitating a balance between objectives and constraints. However, traditional aerodynamic design optimization, particularly in transonic regime, is hindered by its high computational cost. To meet the evolving needs of aircraft design, we present an efficient alternative approach for performing high-fidelity aerodynamic design optimization. The discrete empirical interpolation method (DEIM) is used to obtain a set of interpolation points. These points are utilized to reconstruct flow fields using bases obtained from a database compiled from high-fidelity analyses. The aerodynamic reduced-order model (ROM) is then built based on a Global Local neural network, which introduces a nonlinear mapping function between reduced design variables and the DEIM interpolation points. This data-based ROM offers cost savings compared to the iterative use of Computational Fluid Dynamics (CFD) simulations typically employed in gradient-based design optimisation. The sensitivity analysis is performed on each component that contributes to the development of the ROM. The ROM is utilized to solve the transonic optimisation problem for the undeflected Common Research Model (uCRM). The results obtained using ROM are similar in shape to those obtained using full-order model-based optimisation, with a difference in CD of about two counts. These findings demonstrate the effectiveness of the proposed method for fast and high-fidelity aerodynamic design.

This work presents a strategy to build reduced-order models suitable for aerodynamic shape optimisation, resulting in a multifidelity optimisation framework. A reduced-order model (ROM) based on a discrete empirical interpolation (DEIM) method is employed in lieu of computational fluid dynamics (CFD) solvers for fast, nonlinear, aerodynamic modelling. The DEIM builds a set of interpolation points that allows it to reconstruct the flow fields from sets of basis obtained by proper orthogonal decomposition of a matrix of snapshots. The aerodynamic reduced-order model is completed by introducing a nonlinear mapping function between surface deformation and the DEIM interpolation points. The optimisation problem is managed by a trust region algorithm linking the multiple-fidelity solvers, with each subproblem solved using a gradient-based algorithm. The design space is initially restricted; as the optimisation trajectory evolves, new samples enrich the ROM. The proposed methodology is evaluated using a series of transonic viscous test cases based on wing configurations. Results show that for cases with a moderate number of design variables, the approach proposed is competitive with state-of-the-art gradient-based methods; in addition, the use of trust region methodology mitigates the likelihood of the optimiser converging to, shallower, local minima.

Accurate prediction of turbine rim seal ingestion remains a challenge, even sophisticated unsteady computational models have had limited success due to the complexity and uncertainties present in this type of problems. For this reason, it is of interest to perform stochastic analysis taking into account the variability in the input parameters as well as uncertainties and assumptions associated with a given model of choice. This study focuses on a Secondary Air System (SAS) of a gas turbine, considering a generic cavity in a high pressure turbine (HPT) in which hot gas ingestion occurs and uncertainty in geometrical, operational and modelling parameters is present. Several statistical methods are applied to a 1D gas network model in order to evaluate the impact of the tolerances of the main geometrical parameters in the output, followed by a broader analysis including other operational and modelling variables. Results indicate that most influential parameters are the sealant mass flow rate, a modelling constant and the minimum gap of the wheelspace cavity.