No abstract

In this paper, a systematic CFD work is carried out with the aim to inspect the influence of different cascade parameters on the aerodynamic performance of a reversible fan blade profile. From the obtained results, we derive a meta-model for the aerodynamic properties of this profile. Through RANS simulations of different arrangements in cascades, the aerodynamic performance of airfoils are analyzed as Reynolds number, solidity, pitch angle and angle of attack are varied. The definition of a trial matrix allows the reduction of the minimum number of simulations required. The computed CFD values of lift and drag coefficients, stall margin and the zero-lift angle strongly depend on cascade configuration and differ significantly from standard panel method software predictions. In this work, X-Foil has been used as a benchmark. Particularly, the high influence of pitch angle and solidity is here highlighted, while a less marked dependence from the Reynolds number has been found. Meta-models for lift and drag coefficients have been later derived, and an analysis of variance has improved the models by reducing the number of significant factors. The application of the meta-models to a quasi-3D in-house software for fan performance prediction is also shown. The effectiveness of the derived meta-models is proven through a spanwise comparison of a reversible fan with the X-Foil based and meta-model based versions of the software and 3D fields from a standard CFD simulation. The meta-model improves the software prediction capability, leading to a very low global overestimation of the specific work of the fan.

Near-wall modelling is one of the most challenging aspects of CFD computations. . A compromise between accuracy and speed to solution is usually obtained through the use of wall functions, especially in RANS computations. This approach can be generally considered as robust, however the derivation of wall functions from attached flow boundary layers can mislead to non-physical results in presence of specific flow topologies, e.g. recirculation, or whenever a detailed boundary layer representation is required (e.g. aeroacoustics studies). In this work, a preliminary attempt to create an alternative data-driven wall function is performed, exploiting artificial neural networks (ANNs). The learner that is derived from the multi-layer perceptron ANN, is here used to obtain two-dimensional, turbulent production and dissipation values near the walls. Training examples of the dataset have been initially collected either from LES simulations of significant 2D test cases or have been found in open databases. Assessments on the morphology and the ANN training can be found in the paper. The data-driven wall function is here applied to k-epsilon simulations of a 2D periodic hill with different computational grids and to a modified compressor cascade NACA aerofoil with sinusoidal leading edge. A comparison between ANN enhanced simulations, available data and standard modelization is here performed and reported.

No abstract

Open literature offers a wide canvas of techniques for surrogate-based multi-objective optimization. The large majority of works focus on methodological and theoretical aspects and are applied to simple mathematical functions. The present work aims at defining and assessing surrogate-based techniques used in complex optimization problems pertinent to the aerodynamics of reversible aerofoils. Specifically, it addresses the following questions: how meta-model techniques affect the results of the multi-objective optimization problem, and how these meta-models should be exploited in an optimization test-bed. The multi-objective optimization problem (MOOP) is solved using genetic optimization based on non-dominated sorting genetic algorithm (NSGA)-II. The paper explores the possibility to reduce the computational cost of multi-objective evolutionary algorithms (MOEA) using two different surrogate models (SM): a least square method (LSM), and an artificial neural network (ANN). SMs were tested in two optimization approaches with different levels of computational effort. In the end, the paper provides a critical analysis of the results obtained with the methodologies under scrutiny and the impact of SMs on MOEA. The results demonstrate how surrogate model incorporation into MOEAs influences the effectiveness of the optimization process itself, and establish a methodology for aerodynamic optimization tasks in the fan industry.