PhD Defense - Vinay Prakash

Probabilistic Metamodels to Assess Global NVH Performance of Electric Vehicles

With electromobility comes new challenges in the design and development of passenger vehicles. The noise, vibration and harshness (NVH) attributes of a vehicle are crucial, influencing both passenger comfort and environmental considerations. The rapid evolution of electric vehicles (EVs) necessitates prompt decision-making by NVH designers in the early-stages of the design phase. NVH risk assessments become even more challenging due to the presence of various functional parameters
(e.g., operating conditions, design parameters) and the different uncertainties associated with them for instance, partial knowledge, dispersion in measurement-based data, etc. As a consequence, this work focuses on developing fast and comprehensive probabilistic metamodels able to quantify such uncertainties linking the functional parameters with the global NVH performance indicators. Particular emphasis is placed on assessing narrowband tonal noises originating from electric powertrains in Battery EVs, as well as addressing broadband masking noises resulting from aerodynamic wind and tire-road interaction effects. The chosen methodology employs Bayesian framework coupled with Markov Chain Monte Carlo sampling techniques. This approach facilitates the incorporation of prior knowledge (for instance, coming from automotive domain experts) and enable the propagation of uncertainties across multiple physical domains, which are estimated by semi-analytical approaches further enriched by experimental databases.

The developed probabilistic framework is aimed at providing invaluable support to NVH designers by aiding them to ascertain or refine the functional parameters, assess the global acoustic levels inside the passenger cabin and make informed decisions in the pursuit of optimized vehicle NVH performance.