CEVT is here to make urban mobility enjoyable, smart and sustainable. We fast-track innovation and transform ideas into pioneering technology solutions. Automotive is in our DNA and intelligent mobility is our core. We serve society with sustainable mobility solutions, both private and shared. As part of ZEEKR and the Geely family, one of the most innovative and fastest growing global mobility groups, our work comes to life in products and services from world-leading brands. CEVT was built by people from all over the world. We’re flexible, fast, and always in motion.
Master Thesis Project - Thermal Plant Model Adaptation and Control Software Integration
Subject: Machine Learning, Simulation, Software Development, Thermodynamics
The Energy Management & Controls group is responsible for complete vehicle thermal management strategies. Within the group are technical experts on the air, coolant and refrigerant domains. There is also one simulation part and one SW development part in the group.
Currently the in-house developed thermal management software is tested against requirements using scripts in MIL, SIL and HIL environments. The software is also used together with a very accurate 1-D plant model during GT-SUITE simulations.
Since the GT-SUITE simulations are computationally heavy and time consuming, there is a need for quicker feedback to software developers on how the developed software behaves when coupled with a plant model.
As part of a PhD thesis work, an analytical thermal plant model based on first principle relationships has been developed. This model could be used for quick verification of the thermal management software. Currently the model does not capture some of the highly non-linear dynamics of the real system. This could be improved upon using machine learning methods.
The thermal system dynamics are highly non-linear and too complex to be captured by the first principle modeling. In order to capture the nonlinear dynamics and the system behavior in the corner cases, the existing model should be improved using machine learning methods. The idea is to learn the existing model error based on the test data and model the error using machine learning methods.
A finalized model can be described as
where x(t) is state vector, u(t) is input vector, d(t) is the disturbance, f() is the first principle model and g() is modeled using machine learning to capture the unknow dynamics. This thesis involves formulating the g() model based on Gaussian or neural network (upto student choice). Students are free to investigate other methods to model the nonlinear behavior of the system.
Our in-house developed thermal software is based on Simulink models. Input to the models is primarily sensor readings. Output from the models are primarily actuator requests like compressor speed, valve position etc. This thesis involves coupling the thermal management software models with the analytical plant model, perform simulation and evaluate the software’s basic functionality based on the results.
As a final step, the developed plant model can be integrated in our software tool chain. This would automate its execution and improve the quality of delivered software.
- Improve existing analytical plant model using machine learning methods
- Couple thermal management SW and improved model for co-simulation
- Integrate coupled solution into SW tool chain
- Experience in using Simulink for modelling and simulation
- Knowledge about machine learning methodology
- Basic understanding of thermodynamic concepts such as heat transfer
- Some scripting experience in e.g. Python, Matlab
Number of students: 2, please apply with CV and cover letter.
Starting date: January 2023. The duration of the thesis work is 20 weeks
For more information about the master thesis project please contact:
Mikael Larsson, SW Control Engineer, +46 705 33 74 37, [email protected]
Last application date: 2022-10-28
Apply today. We will perform ongoing selection during the application period. We look forward to hearing from you!
Please note that due to GDPR regulations we can only accept applications sent through the recruitment system, not via email or other channels.