Abstract:This work introduces a new uncertainty quantification method to better deal with scarce data and long simulation run times in Formula 1 design. Race cars are produced in low quantities and for maximum performance. Thus, their designing process is characterised by manufacturing data shortage and complex Computational Fluid Dynamics simulations with long run times. Their car aerodynamics is subject to many random variables that introduce uncertainty into the down-force and drag performance, such as variations in ride height, front wing direction and pitch angle. To accurately predict the car performance during a race, it is important to study the effect of these random variables. This assessment cannot be performed with the standard deterministic Computational Fluid Dynamics approaches used in Formula 1. Even with regard to stochastic approaches, no efficient method has so far been suggested that addresses the problem of data scarcity. The reason for this is that most efficient uncertainty quantification methods fit probability distributions to the scarce data. It is shown in this work that probability distribution fitting can create a significant error using a simple two-dimensional diffuser example. Subsequently, the use of a new data-driven Polynomial Chaos method and its sparse multi-dimensional extension is suggested and demonstrated for Formula 1 to reduce such errors. This method allows to avoid distribution fitting because it is based on pure data. SAMBA’s general formulation also makes it easier to combine any possible inputs within a sparse description for problems with many variables. SAMBA is applied to two realistic car three-dimensional Computational Fluid Dynamics simulations: a NACA 0012 tip wing and the front part of a Formula 1 car. The probabilistic variations of the lift and drag of these two configurations are calculated using SAMBA and shown to be significant.

MULTI-FIDELITY MODELING OF PROBABILISTIC AERODYNAMIC DATABASES FOR USE IN AEROSPACE ENGINEERING

Probabilistic Analysis of Aircraft Using Multi-Fidelity Aerodynamics Databases

A Multi-Fidelity Uncertainty Propagation Model for Multi-Dimensional Correlated Flow Field Responses

Multi-Fidelity Bayesian Neural Network for Uncertainty Quantification in Transonic Aerodynamic Loads

Propagation of Input Uncertainty in Presence of Model-Form Uncertainty: A Multi-fidelity Approach for CFD Applications

Multi-Fidelity Uncertainty Propagation Approach for Multi-Dimensional Correlated Flow Field Responses

ANOVA-based Multi-Fidelity Probabilistic Collocation Method for Uncertainty Quantification

Multi-Fidelity Adaptive Sampling for Surrogate-Based Optimization and Uncertainty Quantification

Data-driven uncertainty quantification for Formula 1: Diffuser, wing tip and front wing variations

Efficient aerodynamic analysis and optimization under uncertainty using multi-fidelity polynomial chaos-Kriging surrogate model

Adjoint Design Optimization Under the Uncertainty Quantification of Reynolds-Averaged Navier-Stokes Turbulence Model

Enhanced multi-fidelity modelling for digital twin and uncertainty quantification

Aerodynamic Data Fusion Towards the Digital Twin Paradigm

A confidence-based aerospace design approach robust to structural turbulence closure uncertainty

Efficient prediction of turbulent flow quantities using a Bayesian hierarchical multifidelity model

Scalable Approach to Uncertainty Quantification and Robust Design of Interconnected Dynamical Systems

Multi-fidelity Deep Learning-based methodology for epistemic uncertainty quantification of turbulence models

Uncertainty Quantification of Artemis I Space Launch System Integrated Aerodynamics Databases

Application of deep learning based multi-fidelity surrogate model to robust aerodynamic design optimization

Multi-fidelity modelling via recursive co-kriging and Gaussian-Markov random fields

Adaptive Multi-index Collocation for Uncertainty Quantification and Sensitivity Analysis