Browse Publications Technical Papers 2023-01-1132

Machine-Learning-Based Modelling of Electric Powertrain Noise Control Treatments 2023-01-1132

Encapsulation of electric powertrains is a booming topic with the electrification of vehicles. It is an efficient way of reducing noise radiated by the machines even in later stages of the design and without altering the electromagnetic performance. However, it is difficult to decide about the optimal treatment. The positions, thicknesses and material compositions need to be optimized within given constraints to reach maximum noise reduction while keeping added mass and cost at minimum. In this paper, a methodology to optimize the encapsulation based on numerical vibro-acoustic simulations is presented. In a first step, the covered areas are identified through post-processing of a finite element acoustic radiation model of the bare powertrain. In a second step, a design of experiment needs to be performed to assess the influence of various cover parameters on the acoustic radiation results. This second step can be very computationally expensive as the number of required virtual experiments increases exponentially with the number of treated regions and parameters for each treated region. In this paper we present a physics-based reduced-order model to overcome this difficulty and do design of experiments in a much more computationally affordable manner. It is then enriched with machine learning to provide finer tuning of the treatments that we place on the target regions. This would allow the final designer to iterate between treatment strategy in the matter of seconds, paving the road for an advanced optimization algorithm. The accuracy of the presented model is also detailed.


Subscribers can view annotate, and download all of SAE's content. Learn More »


Attention: This item is not yet published. Pre-Order to be notified, via email, when it becomes available.
Members save up to 17% off list price.
Login to see discount.
Special Offer: Download multiple Technical Papers each year? TechSelect is a cost-effective subscription option to select and download 12-100 full-text Technical Papers per year. Find more information here.