The automotive industry is in the middle of a turbulent shift from personally owned, gas powered cars to autonomous, shared, electric mobility solutions such as robot axis. This shift changes key assumptions in automotive design: usage cost becomes much more important and all the use cases that a single owned vehicle had to support can now be covered by a fleet of shared vehicles. This change in stakeholder presents opportunities for new system architectures for vehicles. Current optimization techniques often focus on the component level, while the biggest impact is often to be found on the systems’ level. The systems engineering process used in industry to derive architectures is often tedious and informal, resulting in suboptimal architectures. Formalizing and partially automating part of the concept design phase to enable fast creation and evaluation of architectures can lead to more optimal system architectures for vehicles.
In this context, the topology of a system, namely the choice, placement and interconnection of its components, has a significant impact on its achievable performance and must be carefully studied. Therefore, the possibility to rapidly generate automatically optimized system architectures from a set of desired functionalities and a library of component technologies (platform) would ultimately greatly accelerate the deployment of electric vehicles and lead to better performance on the cost function (typically total-cost-of-ownership). The key questions relate to how to automatically extract engineering knowledge that supports the top-down design process by understanding which subsystems or components from the platform fulfil certain system-level functional requirements and how to connect them by formulating mathematical constraints as in a bottom-up design process. This research aims at devising discrete mathematical models and optimization models and methods for system-level powertrain design. Specifically, the student will first investigate state-of-the-art discrete modelling tools. Second, methods will be studied by using machine learning to translate functional requirements and engineering models to a general constrained optimization setting. The student will leverage and advance optimization methods to solve such discrete optimization problems and devise a platform-based design toolbox to perform extensive case-studies for a family of electric vehicles. The research concentrates on the transition from requirements to embodiment-in-design in the early conceptional design phase by automation that is supported by machine learning, constraint programming and model-based engineering methods.
The PhD project (part of work package 4: Electric Mobility) is performed within the framework of a larger NWO research project NEON that is funded on the support of the acceleration of the transition towards sustainable energy and mobility.Functie-eisen