Conference Agenda
Overview and details of the sessions of this conference. Please select a date or location to show only sessions at that day or location. Please select a single session for detailed view (with abstracts and downloads if available).
|
|
|
Session Overview |
| Session | ||||
WO3: Oral Session WO3 - Machine Learning #2
| ||||
| Presentations | ||||
Machine Learning-based Reduced Order Modelling of Nonlinear Magnetic Devices 1Department of Industrial Engineering, University of Padova, Via Gradenigo 6/A, 35131, Padova, Italy; 2Department of Information Engineering, University of Padova, Via Gradenigo 6/B, 35131, Padova, Italy; 3Department of Electrical, Computer, and Biomedical Engineering, University of Pavia, Pavia 27100, Italy This work presents an approach to efficiently model nonlinear magnetic problems through a machine-learning approach. Proper Orthogonal Decomposition (POD) is employed to reduce the model's dimensionality while machine-learning techniques are employed to find a relationship between the input data and the solution. In this digest, the methodology is demonstrated on a simplified version of the TEAM 36 benchmark problem, involving the induction heating of a steel billet with nonlinear magnetic properties. Results show that the approach accurately captures nonlinear effects while significantly reducing computational demands, with future work focusing on extending it to a more realistic representation of the device.
Data-Driven Modeling of Accelerator Magnets CERN, Switzerland We present a methodology that enables data-driven modeling and traceable design workflows of accelerator magnets by combining numerical field simulations with tests and magnetic measurements. The aim is to create models updated by magnetic measurements that allow the extrapolation of performance parameters regarding different magnet powering cycles, manufacturing defects, and material parameters while mitigating the uncertainties in both the simulations and measurements.
Tensor Train-Based Approach for the Magnetic Moment Method 1LAPLACE, Université de Toulouse, CNRS, INPT, UPS, Toulouse, France; 2Inria centre at university of Bordeaux In this work, we focus on applying the Magnetic Moment Method (MMM) and enhancing its computational efficiency using a low-rank Tensor Train (TT) approach. By transforming the dense linear system generated by MMM into TT format, we achieve significant reductions in both computational cost and memory usage. This method is tested on several problems involving a ferromagnetic part with a regular mesh, demonstrating superior performance compared to traditional techniques.
Data-driven Topology Optimization with Warm Starting CMA-ES for Constraint Optimization of Rotating Machine 1Department of Electrical, Electronics and Information Engineering, Nagaoka University of Technology; 2Muroran Institute of Technology; 3Faculty of Science and Engineering, Kindai University In this paper, a method, which reduce the computational cost for solving constrained optimization, was proposed. In constrained optimization problems, it takes a lot of time to search for a solution that satisfies the constraints function, and the computation time required to improve the objective function increases. Therefore, this study proposes a method based on Warm Starting CMA-ES. By using the proposed method, solutions which satisfy the constraint function can be found quickly, thus reducing computational time for optimization. To validate the effectiveness, the proposed method was applied to the topology optimization problem of interior permanent magnet motor.
| ||||