Master Thesis – Reduced Order Modeling

Master Thesis – Reduced Order Modeling

Introduction


Developing control system software for the thermal management functionality of a vehicle is becoming more and more complex. This process involves, amongst other things, running many simulations of the complete thermal system of the vehicle in conjunction with the control system prototype software. These simulations are very complex, and the software development could be sped up significantly if the execution time could be improved. The focus of this thesis will be to develop a methodology to replace these simulation models with a machine learning model of acceptable accuracy with a significant improvement in execution time.
Who we are:
The Thermal Management department within Vehicle Technology is responsible for developing, delivering, and maintaining an optimized Cab Climate and Thermal Supply systems for all types of propulsion installations to all truck brands within the Volvo Group. We are responsible for leading the work with strategies and advanced engineering globally. We are located at Gothenburg and Bangalore, and we have close cooperation with the engineering sites located in Greensboro and Lyon. We understand the final customer needs and apply our knowledge to develop technical concepts and solutions that satisfy customer and business needs. The work is based on innovation, shared technology, common architecture and brand uniqueness. We want to make a difference by being there for our customers and by providing uptime and reliable products.
As a master thesis student in the Thermal Management Verification and Validation team you will be a valued contributor to our deliveries and continuous learning. You will be surrounded by a global and diverse team of highly skilled and engaged colleagues who will be interested in the progress of your work and eager to help and support along the way.


Description of tasks and expected outcome


Literature study on what type of machine learning technique would be most appropriate for this type of application
Investigate how to create training data in the best way using GT-Suite
Use the training data to train promising machine learning models, and compare their effectiveness, as well as a comparison to the existing GT-suite simulation model.
Document best practice for data generation, as well as the process for choosing an appropriate machine learning model


Suitable background


This master thesis requires analytical skills and a good understanding of thermodynamics, data handling, as well as machine learning techniques. To be successful in this master thesis project we believe that it is important that you recognize yourself in the following description.
Final year student in Master program for Automotive, Physics, Computer Science, Mechatronics or similar
Experience with machine learning frameworks like Tensorflow or Pytorch are a merit
Programming skills in Python, Matlab or similar
Fundamental understanding of thermodynamics
Fundamental knowledge of vehicle technology
Analytical mindset and problem-solving skills
Fluent in English

Thesis level: Master, 30 ECTS credits


Number of students: 2


Start date: Second half of January 2024, or upon agreement


Industrial supervisors: Marcel Aarts and Dongyu Liu


Location: Volvo Lundby site, Gothenburg, Sweden


We look forward to receiving your application!


Kindly note that due to GDPR, we will not accept applications via mail. Please use our career site.