30 HP - Evaluation of deep learning methods for sound source localization in a powertrain test bench

Introduction
Thesis work is an excellent way to get closer to Scania and build relationships for the future. Many of today's employees began their Scania career with their degree project.

Background 
Acoustic characterization of powertrains is essential to designing better trucks and buses from the point of view of driver comfort, fulfilment of legal requirements and overall customer perception. The two most relevant acoustic characteristics of a powertrain are spectral content and directivity, the knowledge of which can then be used to localize various noise sources in the powertrain. While measuring the spectral content in an acoustic test bench is a relatively straightforward, determining its directivity particularly at higher frequencies is complicated. The primary challenge is the microphone array, which is usually located in the far-field of the test object and is quite sparsely distributed for the frequencies of interest and the dimensions of the test object.

Objective & Job Description
The goal of the thesis is to explore the potential and advantages of deep learning approaches using Convolutional Neural Networks (CNN) – as an alternative to classical signal processing and matrix inversion based approaches – to solve the problem of sound source localization from sparse far-field microphone measurements. These measurements are performed in a semi-anechoic powertrain test-bench.

Some important features of the model that will be explored in detail are the resolution of the velocity/pressure distribution in the near-field of the test object, frequency range of validity, neural network architecture and training/validation dataset generation. For the purposes of dataset generation, both FEM simulations and measurements will be explored.

Education/program/focus
The student must have completed at least 60 ECTS credits in a Master’s degree program belonging to any one of the below programs or other related programs:

• Mechanical/civil/aerospace engineering, technical physics, applied mathematics and computer science
• Experience with ML and Python packages is preferred.
• Knowledge of acoustics and signal processing is a bonus.
• Fluent in English, written and spoken.
• The duration of the thesis work is expected to be between 20-25 weeks at will be carried out onsite at Scania’s R&D facilities in Södertälje, Sweden starting in January-February 2025.

Contact persons and supervisors
Supervisor: Dayasagar Valady Srinivasan, Development Engineer Acoutics, ENTAK
dayasagar.srinivasan@scania.com
+46 855 350 802   
Manager: Joakim Lindholm, Head of Gear Technology & NVH, ENTAK
joakim.lindholm@scania.com
+46 855 351 280

Application:
Your application must include a CV, personal letter and transcript of grades 

A background check might be conducted for this position. We are conducting interviews continuously and may close the recruitment earlier than the date specified.     

 

Thesis work within Acoustics & AI