The CRC is organised in four scientific Project Areas, as well as an area of central projects for research training, research data management and project management. The four scientific areas are classified into electromagnetism, fluid mechanics, numerics and optimisation. All projects are interwoven, with the regular exchange of results, including models and algorithms. Validation is ensured by integrated measurements. The different multiphysical and multiscale aspects are tackled in a holistic approach with a project plan designed to fully address the interdisciplinary character of the task.
Project area A: Electromagnetics
Modern electric machines typically operate within a wide range of torque-speed operating points. To evaluate the overall performance of a given drive over a given drive cycle, time-consuming model-based time-stepping analyses or experimental investigations of the cycle-over-time can be carried out. Alternatively, the performance can be evaluated based on torque-speed performance maps that provide the steady-state performances for the individual operating points. This project will address the fundamental questions on the accuracy of the use of such performance maps, and expand into alternative approaches drawing from the collaboration within the CRC.
This project enables electric machine simulation using a domain decomposition method based on isogeometric analysis for electromagnetic problems. Non-conforming trimmed interface discretisations are glued to each other by mortaring. Parallelised subdomain-wise model order reduction speeds up repeated computations such that uncertainty quantification and optimisation become feasible. Finally, the reduced order models will be used as coarse propagators within the parallel-in-time method Parareal.
This project will enable transient machine simulation with the same predictive power as standard 3D finite-element models, but at a computation cost reduced from several hours to a few minutes. This will be achieved by error-controlled adaptation in space and time and by tailored hybrid modelling techniques. The adaptive resolution will be exploited to construct high- and low-fidelity models which, combined with an appropriate model management, will accelerate outer-loop algorithms such as, e.g., optimisation.
Insulation systems in electric machines limit the permissible operating range and significantly determine the service life. Therefore, a complete knowledge of the electrothermal insulation system problem is essential for the mission of the CRC of developing holistic simulation design routines. This project models machine insulation systems as a coupled electric and thermal problem including realistic variabilities of the insulation material parameters and geometry. Transient or cycling operating stresses are simulated to predict insulation failures and to assess novel cooling strategies.
To increase the efficiency of electrical machines, it is essential to comprehend the magnetic loss mechanism and its dependency on the local defects and grain structure of the magnetic core materials. In this project, we will apply the micromagnetics model based on the Landau-Lifshitz-Gilbert equation, computational homogenisation and Machine Learning to calculate macroscopic hysteresis and losses, which are informed by both local defect features at grain boundary and by microstructural grain structure. Based on simulation data of a large number of local structure samples and grain structure samples, which are either measured or synthetically generated, surrogate models for the linkage between local features and reverse field and for prediction of macroscopic hysteresis and loss will be also obtained by Machine Learning. The multiscale and data scheme will be validated via experimental data and applied for case studies related to cutting and aging.
Cooling is essential for light-weight electrical machine development, in particular for unsteady machine operation. Heat can be most efficiently transferred by multi-phase fluid flows. The present project investigates rotor cooling by heat pipes, which are attractive for both efficiency and simple design. Optimisation potential by the working medium and the structure of a porous medium inside, together with system rotation, is investigated. The project develops a highly accurate numerical model using an eXtended Discontinuous Galerkin method to calculate the coupled flow and heat transfer in heat pipes in detail.
Overheating of the rotor in electric motors may alter the magnetic properties of rotor-integrated surface magnets or damage electric insulation layers. The present study aims to significantly enhance current airflow based direct surface cooling principles by adding small liquid drops to the flow passing through the rotor-stator gap. The interaction between airflow flow, droplet dynamics and substrate properties on the resulting heat flux will be studied in detail in a generic rotor test section. The aim of the planned research is to experimentally identify driving cooling mechanisms to develop an experimentally validated model for aerosol based thermal load peak treatment in a rotor-stator gap.
This project deals with the analysis and time domain simulation of the heterogeneous coupled system of differential algebraic equations arising from the multiphysical description of an electric machine. A system-level approach allows to couple existing models for the different physical subsystems that are involved and successively improve their predictive capability by including improved models that arise from other projects. The aim is to allow for a full drive cycle simulation by means of efficient multirate methods as well as parallelisation in space and time.
This project addresses the systematic derivation and analysis of coupled field models describing the complex physical behaviour of electric machines and their components during transient operation. An energy-based variational paradigm is used, in which the energy-dissipation of the resulting coupled nonlinear dynamical systems is incorporated explicitly through the specific form of the mathematical models. Approriate structure-preserving discretisation strategies are devised and analysed for the reliable and accurate simulation in the presence of nonlinear material response, strongly varying length and time scales, complex and moving geometries as well as various kinds of uncertainties.
An accurate geometric representation is an absolute must for reliable computational simulations, especially for electrical drives where the air gap between the stator and rotor is crucial. The project will use the isogeometric paradigm combined with level-set functions to enable an advanced discretisation framework that addresses this challenge. Thereby, it provides fundamental routines for the novel space-time and optimisation methods developed in other projects of this CRC. Finally, the exchange and subsequent editing of the optimised shapes is enabled by generating analysis-suitable boundary representations compatible with current CAD standards.
An efficient, reliable, fast, parallel and accurate direct numerical simulation tool for the solution of time-dependent partial differential equations is a mandatory ingredient for the design and optimisation for electrical machines. Space-time finite element methods and related domain decomposition methods allow an adaptive resolution and parallelisation strategies simultaneously in space and time. Applications involve general parabolic evolution equations, and in particular the eddy current approximation of the time-dependent Maxwell equations.
Noise and vibration (NVH) of electric drives is one of the design criteria for effective modern designs. Due to the broad band of exciting frequencies a time domain formulation for the simulation of structural vibrations and sound will be developed. The focus of the project will be on the efficient realisation of an IGA based boundary element method in time-domain, including variable time step sizes (generalized convolution quadrature method) and data sparse representations (generalized adaptive cross approximation) with respect to the spatial and temporal variables.
This project will develop data-driven surrogate modelling methods to enable uncertainty quantification (UQ) studies within the context of electric machine design. First, material and geometrical uncertainties in the form of random fields will be stochastically modelled. Next, machine learning regression algorithms will be designed to approximate the dependence of electric machine quantities of interest on uncertain design parameters. Last, surrogate-based UQ methods for failure probability estimation and multivariate sensitivity analysis, two important and computationally demanding UQ tasks, will be developed to quantify the impact of uncertainty, provide novel insights, and facilitate improved machine designs.
This project deals with the topology optimisation of electric machines considering not only electromagnetic fields, but also thermal fields and their interaction. The underlying problem is modelled as a system of nonlinear, time-dependent partial differential equations that is considered in both two and three space dimensions. The optimisation is carried out by means of a multi-material level set algorithm based on the topological derivative and takes into account also practical feasibility and mechanical stability of the designs as well as multiple performance criteria.
Production tolerances, material variations, usage scenarios and other influences lead to uncertainty in the optimal design of electric machines. The first objective of the project is the development, analysis and application of robust optimisation methods for geometry and topology optimisation of electric machines under uncertainty. By using first and second order approximation techniques we plan to develop (distributionally) robust optimisation formulations and efficient adjoint-based solvers invoking reduced order models and error estimators. Moreover, we plan to develop optimisation methods for the optimal design of experiments such that resulting estimates of material parameters by solving inverse problems have minimal variance.
For the design of new electric motors, the precise knowledge of local magnetic properties and a hysteresis model for accurately estimating losses is of utmost importance. The goal of the project is the development of a combined method based on measurements, 3d simulations and inverse schemes to locally determine the magnetic properties of electric steel sheets and iron cores as used in electric machines. In particular, we will be able to determine, e.g., the change of the magnetic properties of ferromagnetic materials after being cut, punched, welded, etc.
The Research Data Management project will establish software technologies to support data exchange between all subprojects of the SFB. This includes the exchange between experiments and numerical simulations as well as the exchange between different kinds of simulations. For these purpose, central repositories for code and data will be set up, thus creating a digital representation of the entire CRC. The integrity of these repositories will be ensures by applying Continuous Integration techniques. Furthermore, the project it is the central contact for CRC members in questions of research data management.
We would like to customise the information and usability of this website to your preferences and needs.
To this end, we use so-called cookies. Please choose which cookies you would like to enable when visiting our webpages.
Some of these cookies are required to load and correctly display this website on your device.
These are strictly necessary or essential cookies and cannot be deselected.
The preferences cookie saves your language setting, while the statistics cookie regulates
how the open-source statistical software “Matomo” analyses your visits to and activities on our website.
For more information about cookies we use, please refer to our