Adaptive RBF with hyperparameter optimisation for aeroacoustic applications

Abstract

The present work reports an investigation on the use of adaptive metamodels based on radial basis functions (RBFs) for aeroacoustic applications of highly innovative configurations. The relevance of the topic lies on the paramount importance of metamodelling techniques within the design optimisation process of disruptive aircraft layouts. Indeed, the air traffic growth, consequently the hard environmental constraints imposed by regulations, will make a technological breakthrough, an imperative need within few years. As a consequence, the engineering community is paying particular attention to the development of innovative techniques for the design of unconventional configurations. For this class of applications, the designer cannot successfully rely on historical data or low-fidelity models, and the expensive direct simulations remain the only valuable design strategy. In this regard, it can be demonstrated that the use of surrogate models, i.e., metamodels, significantly reduces the computing costs, especially in view of a robust approach to the optimised design. In order to further improve the efficiency of metamodel-based techniques, dynamic approaches based on hyperparameter optimisation and adaptive sampling procedures have been recently developed. The case study presented here pertains the exploiting of dynamic RBF-based metamodels for noise shielding applications. The analysis of the metamodel performances and its convergence properties shows how the final number of direct simulations is significantly reduced by the hyperparameter optimisation algorithm, still strongly depending on the choice of the RBF kernel function.

Experimental characterisation and data-driven modelling of unsteady wall pressure fields induced by a supersonic jet over a tangential flat plate

“…The dichotomy between biological neurons and mathematical processes has been investigated in the 1940s (McCulloch & Pitts 1943;Hebb 1949), but the theoretical layout of complex multilayer structures was developed within the last 50 years (Ivakhnenko 1967(Ivakhnenko , 1973 thanks to the increase in computing resources. The ANNs are widely used in computer vision and speech recognition (Safran & Shamir 2017) and, more recently, are also being employed in the construction of surrogate models in the aeroacoustic field, alongside the well-assessed radial basis functions dynamic models (Centracchio et al 2021b;Burghignoli et al 2022).…”

Subsonic installed Jet noise scattering prediction using BEM and near-field-data-optimized Wave-Packet model

“…There is hence a strong need for low-and mid-fidelity models and solvers able to catch the fundamental feature of installed jet noise, avoiding the solution of the complete set of equations holding the dynamic of the complex fluid structures involved. Adaptive metamodeling techniques have been recently applied to this class of problems [19,20] to reduce the computational effort required in determining the optimal position of the propulsion system that minimizes the noise directed towards the ground and community. Artificial Neural Networks, trained with experimental data in both near and far field, have been used as a non-linear surrogate model to predict the noise emitted by a single-stream jet in under-expanded conditions [21], showing good agreement for a wide range of Mach number.…”

Data-Driven Multiobjective Optimization of Wave-Packets for Near-Field Subsonic Jet Noise

“…There is, hence, a strong need for fast models for predicting the effects of the propulsion system installation on the acoustic emissions at the aircraft level, enabling its assessment from the first design stages and also for disruptive configurations. Adaptive metamodeling techniques have been recently applied to this class of problems [16,17] to reduce the computational effort required in determining the optimal position of the propulsion system that minimises the noise directed towards the ground and community. Boundary Elements Method (BEM) simulations involving the monopole as a noise source are often used to feed the model creation.…”