ABSTRACT

We consider the vehicle platoon problem from an aerodynamic point of view. That is, the multiparametric problem of predicting the aerodynamic forces acting on a vehicle following another one under realistic road conditions. We develop a RANS methodology to simulate accurately the physics governing the problem and devise a multifidelity methodology to approximate the problem using as few computational resources as possible. In the first part of the thesis we develop and analyse the surrogate methodology in the framework of the Smolyak approximation method. We extend the Multi-Index Stochastic Collocation (MISC) method to handle the problem in a more efficient way. This includes the possibility of constructing a single surrogate for vector-valued functions as well as the use of a parametric domain with categorical variables representing a finite number of possible leading vehicle geometries.

For this, we formulate the Smolyak approximation in a more abstract way and implement it in a modular C++ code using efficient and stable numerical algorithms. We also provide some novel convergence bounds of the method and validate them with numerical examples. We also address the problem of unstability that manifests in some problems where MISC has been applied in the form of spurious oscillations. We identify the reasons of the oscillations as irregularities present in the low fidelity data and prove a theorem in the general framework of multifidelity that shows that such irregularities must be avoided to ensure convergence of the surrogate.In the second part of the thesis we study different efficient CFD methodologies to simulate accurately the flow past a vehicle in realistic road conditions. This includes the presence of cross-wind and the disturbances of the flow due to other vehicles in typical platoon and overtake manoeuvres. Finally we apply the extended MISC method to a multiparametric platoon problem to construct a surrogate of the three components of the aerodynamic force acting on a vehicle under realistic platoon conditions.

Committee

• President: Dr. Elias Cueto
• Secretary: Dr. Pedro Diez
• Member: Dr. Fabiola Cavaliere

PhD Advisors

• Dr. Antonio Huerta
• Dr. Sergio Zlotnik

PHD CANDIDATE

Mr. Rafel Perelló i Ribas is a PhD candidate in Civil Engineering at CIMNE’s Credible High-Fidelity and Data-Driven Models research cluster. He holds a MSc in Numerical Methods in Engineering from UPC and Bachelor’s Degree in Industrial Technology Engineering from UPC. His research covers multifidelity surrogate modelling and computational fluid dynamics with applications to ground vehicle aerodynamics.