Magnetic losses under triangular symmetric and asymmetric induction waveforms, like the ones found in power electronics devices, have been measured over a broad range of frequencies and predicted starting from standard results obtained with sinusoidal induction. Non-oriented Fe-Si and Fe-Co sheets, nanocrystalline Finemet ribbons, and Mn-Zn ferrites have been investigated up to f = 1 MHz and duty cycles ranging between 0.5/f and 0.1/f. The intrinsic shortcomings of the popular approach to loss calculation of inductive components in power electronics, based on the empirical Steinmetz equation and its numerous modified versions, are overcome by generalized application of the Statistical Theory of Losses and the related concept of loss separation. While showing that this concept applies both to ferrites and metallic alloys and extracting the hysteresis (quasi-static), excess, and classical loss components, we relate in a simple way the magnetic energy losses under symmetric triangular induction (square wave voltage) and sinusoidal induction. The loss behavior under asymmetric triangular induction is directly retrieved from the symmetric one, by averaging the energy losses pertaining to the two different semi-periods. Good comparison with the experimentally measured energy loss versus frequency behavior is verified in all materials.

A new axial flux permanent magnet machine (AFPM) with soft magnetic composited (SMC) cores is proposed for electric vehicle (EV) application in this paper. As its windings are wound on the stator ring core, it can be regarded as a toroidally wound internal stator (TORUS) machine. With the adopted SMC material for stator core, this machine has the benefits of 3-D magnetic flux properties. The windings and SMC cores can be designed to form a very compact structure, and thus, the torque density can be improved greatly. To obtain the a good flux concentrating ability, two TORUS machines are designed and analyzed, one is with NdFeB magnet for high-performance EV application and the other is with the cheap ferrite magnet for low-cost application. The 3-D finite-element method is used to analyze the electromagnetic parameter and performance. For performance comparison, a commercial AFPM with yokeless and segmented armature P400 is used.

Compared with traditional induction machine and direct current machine, permanent magnet synchronous machine (PMSM) holds many merits like higher torque ability and efficiency when high magnetic co-energy sintered NdFeB magnet is used. However, for the operation with high frequency, the resulted eddy current loss by the permanent magnet (PM) is very high and this kind of loss can bring the PM with high-temperature rise, making the PM face the risk of irreversible demagnetization. To reduce the PM eddy current loss, complete magnet segmentation is an effective method. However, taking this kind of method will increase manufacturing cost and reduce the mechanical robustness of the PMSM. Thus, a partial magnet segmentation method was proposed in the past. In this paper, a new annular partial segmentation (APS) method is proposed for the reduction of the PM eddy current loss, including single-side APS and double-side APS configurations. Considering that the additional process on the PM will reduce the mechanical robustness of the PM and the electromagnetic performance of machine, both the electromagnetic performance and the mechanical strength of the PM have been analyzed, based on 3-D finite-element method. It can be found that if the proposed new annular partial segmentation (APS) method is adopted, the eddy current loss in the PM can be reduced greatly, while the mechanical robustness of the PM can be guaranteed comparing with the traditional partial magnet segmentation method.

Power transformer is the main part of the power grid. After some tests and operations of transformers, a certain residual flux density ( $B_{r}$ ) is left in the iron core due to the ferromagnetic material hysteresis phenomenon. When the transformer is switched on with no load, $B_{r}$ will lead to inrush current. Therefore, research on measuring $B_{r}$ is of great significance for transformer protection. In this article, a $B_{r}$ measurement method based on applied dc voltage is proposed. First, combined with the magnetic circuit and transient circuit analysis, the relationship between $B_{r}$ and the extremum value of the transient current difference is obtained. Then, the finite element method (FEM) is used to calculate the transient current under different $B_{r}$ and test circuit parameters; the suitable test circuit parameters will be determined, and then the empirical formula for calculating $B_{r}$ is obtained. Finally, the experimental results verify the correctness of the simulation analysis.

When the no-load transformer is connected to the power grid, an inrush current may be resulted as the existed residual flux in the transformer core. To suppress this inrush current, the detail in-formation of the residual flux needs to be known. In this paper, a transient current method for obtaining the residual flux density in the single-phase transformer is proposed. Firstly, two DC volt-ages with opposite polarities are sequentially applied in the meas-uring winding, and then two corresponding transient response currents can be detected. Based on above information, an empiri-cal equation in regarding of the residual flux density and the transient current difference will be obtained. To make sure the in-itial residual flux density does not change greatly, the magnitude of the DC voltage has been determined as well. Finally, the pro-posed method is verified by the experiment results.

The objective of this study was to explore the potential of machine-learning techniques in the automatic identification and classification of brain metastases from a radiomic perspective, aiming to improve the accuracy of tumor volume assessment for radiotherapy. By using various machine-learning algorithms, including random forest, support vector machine, gradient boosting machine, XGBoost, decision tree, artificial neural network, k-nearest neighbors, LightGBM, and CatBoost algorithms, a stacking ensemble model was developed to classify gross tumor volume (GTV), brainstem, and normal brain tissue based on radiomic features. Multiple evaluation metrics, including the specificity, sensitivity, negative predictive value, positive predictive value, accuracy, Matthews correlation coefficient, and the Youden index, were used to assess the model’s performance. The stacked ensemble model integrated the strengths of the nine base models and consistently outperformed individual base models in classifying GTV (area under the curve [AUC] = 0.928), brainstem (AUC = 0.932), and normal brain tissue (AUC = 0.942). Among the base models, the support vector machine model demonstrated the best performance in the three classifications (AUC = 0.922, 0.909, and 0.928). The higher performance of the stacked ensemble model highlighted the low performance of other models, including the decision tree (AUC = 0.709, 0.706, 0.804) and k-nearest neighbors (AUC = 0.721, 0.663, 0.729) models in certain contexts, such as when faced with high-dimensional feature spaces. While machine learning shows significant promise in medical image analysis, relying solely on a single model may lead to suboptimal results. By combining the strengths of various algorithms, the stacking ensemble model offers a better solution for the classification of brain metastases based on radiomic features.