Thesis work 30hp - Hybrid model development for real time simulation

Thesis work 30hp - Hybrid model development for real time simulation

Ingress:

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

Background:

With the increasing demand on virtual environments for software testing and prototyping, we must support and simulate more systems. Some of these systems are straightforward to describe using conventional differential equations. However, simplified equations are usually used for simulation to meet the real time requirement.

One of the key challenges is capturing additional dynamics based on measurement data to improve simulation reliability while maintaining real time constraint.

Target:

The goal of this thesis is to investigate the applicability and limitations of hybrid physics-based and data-driven modeling for simulation of dynamical systems.

The central research question is under which conditions a simplified physical model can successfully be complemented with data-driven methods to construct accurate, robust, and computationally efficient simulations. In particular, we will provide relevant powertrain-related systems to be examined and compare with existing baselines.

Example assignments:

• What approaches are most suitable for combining physics-based equations with ML? Subspace/system-identification-based approaches, Ensemble methods such as XGBoost or similar are some examples.

• How much can the residual error y-y^ be reduced by using ML to compensate for the limitations of a simplified physical model?

• Under what conditions, and using what methods, can we ensure that the resulting hybrid model generalizes well to unseen operating conditions?

• What data and input signals are required to complement the physical model?

• Can we systematically determine which input variables are necessary to achieve sufficient model accuracy?

Education: MSc in Applied Mathematics or Engineering with specialization in Control Systems, Engineering Physics, Signal Processing or Computer Science

Number of students: 1.