Conference Agenda

The effect of dynamic mechanical stress on the magnetic behavior is modeled by the combination of a vector-play model and a simplified multiscale approach. The parameters for the dissipative piezomagnetic behavior are identified using the same methodology as in the purely magnetic case, making use of a set of measured coercive fields. The model is shown to reasonably capture the material response under dynamic magneto-elastic loadings.

In this work we present an approach for the parameter identification of an energy based hysteresis

model from measurement data. To identify the parameters a neural network is trained to predict

the needed hysteresis model parameters for given measurement data. The hysteresis model is implemented as it would be fully

differentiable and is used for the training of the neural network. The trained network can then be used to

identify the parameters of the hysteresis model without retraining. To achieve this, synthetic

training data is used and the approach is validated on measurement data. Furthermore, we investigate the

performance of the proposed pipeline with and without the differentiable hysteresis model at its core.

The linear and passive devices with distributed parameters, are modelled as multi-port Hamiltonian (pH) systems with a finite number of ports, coupled to external structures with lumped parameters. Appropriate boundary conditions (BC) for the Partial Differential Equations (PDEs) of several physical fields inside devices are used. Originally, they are Electric Circuit Element (ECE) BC, but they are generalized for multi-disciplinary domains, such as elastic solids, and acoustic and thermal devices, having two scalar interaction variables: the flow and effort for each terminal. Their internal field is discretized by the Partitioned Finite Element Method (PFEM), using Lagrange multipliers. So, for several devices which may have ECE BC, Absorbing Boundary Conditions and/or Classical BC for incident tangential field Et are generated in Differential Algebraic Equations (DAE) port-Hamiltonian canonic form.

The magnetization and loss characteristics of the electrical steel sheet under DC bias are complex, especially since the hysteresis loss is highly nonlinear. Many mathematical fitting methods for the loss prediction of electrical steel sheets require a large number of data samples, so it is not suitable for engineering loss calculation under DC bias. The Tellinen model is an efficient and flexible model that can describe any scalar magnetic process, and only requests one branch of the limiting hysteresis loop to simulate static hysteresis loops. Therefore, based on the variation range of the AC-DC superimposed magnetic field under DC bias, the asymmetric quasi-static hysteresis loop can be simulated to calculate the nonlinear hysteresis loss. By using the statistical theory of losses, the calculation methods of classical loss and excess loss under DC bias are analyzed. Through the superposition of hysteresis, classical, and excess losses, a novel loss calculation method of the electrical steel sheet under DC bias is achieved, and the simulation results are in good agreement with the experimental measurement results. Also, the loss predicted of electrical steel sheets under DC bias can be carried out in this method.

The winding eddy current loss can cause a dangerous temperature rise in an electric machine, especially in a coreless topology. The coreless axial-flux machine is an unconventional machine configuration with many advantages, but its application is limited due to the winding eddy current loss. However, this problem has not yet been studied extensively in previous literature due to the relatively new topology. Furthermore, it is very difficult to analyze this loss accurately, because it would usually require modeling each of the machine’s conductors, extending the designing and computation time significantly. This article is one of the few that utilize a hybrid method to calculate winding ac loss with both analytical method and simplified 3D finite element analysis of a coreless axial-flux machine, under both no load and load conditions. This method can reduce the computation time while giving relatively good accuracy, making the coreless axial-flux topology becomes more feasible. Losses at multiple sizes of conductors were calculated using this method and thermal analysis was performed with 3D FEA to determine which conductor size will not cause the thermal problem. The analysis results of losses and temperature are then validated with test results.