Interconnect parasitic capacitance extraction is crucial in analyzing VLSI circuits’ delay and crosstalk. This paper uses the deep neural network (DNN) to predict the parasitic capacitance matrix of a two-dimensional pattern. To save the DNN training time, the neural network’s output includes only coupling capacitances in the matrix, and total capacitances are obtained by summing corresponding predicted coupling capacitances. In this way, we can obtain coupling and total capacitances simultaneously using a single neural network. Moreover, we introduce a mirror flip method to augment the datasets computed by the finite element method (FEM), which doubles the dataset size and reduces data preparation efforts. Then, we compare the prediction accuracy of DNN with another neural network ResNet. The result shows that DNN performs better in this case. Moreover, to verify our method’s efficiency, the total capacitances calculated from the trained DNN are compared with the network (named DNN-2) that takes the total capacitance as an extra output. The results show that the prediction accuracy of the two methods is very close, indicating that our method is reliable and can save the training workload for the total capacitance. Finally, a solving efficiency comparison shows that the average computation time of the trained DNN for one case is not more than 2% of that of FEM.
15 September 2022 in IEEE Transactions on Microwave Theory and Techniques
This article proposes a method to engineer degenerate band-edge conditions in periodic substrate-integrated waveguides (SIWs). A degenerate band edge (DBE) is a fourth-order exceptional point originating from the coupling of four Floquet–Bloch modes; it leads to a dispersion relation flatter than a regular band edge (RBE) and is suitable to design slow wave structures and high $Q$ -factor resonators. Design guidelines to achieve a DBE point are given based on Bloch analyses of unit cells. An oblique line of vias in SIW is proposed to form a suitable modal coupling between adjacent waveguides enabling a degenerate band-edge condition. The effect of losses and the robustness to parameter variations are discussed by means of rigorous Bloch analyses. A unit cell showing the best behavior concerning the losses and geometric tolerances is selected and prototypes of finite-length resonators are realized. Measurements validate simulated results and confirm the typical $N^{5}$ scaling of the loaded $Q$ -factor to the number of cells $N$ of the resonator.
10 November 2025 in IEEE Transactions on Magnetics
The previous work on the demagnetization fault detection of permanent synchronous magnet motor (PMSM) focuses on giving information of the fault’s existence or severity. However, the magnetic field distribution is rarely considered which provides the detailed model of the machine. As the sensor configuration affects the accuracy of the reconstructed field, in this paper, a greedy algorithm associated with a reduced order model based on POD is proposed to solve the sensor placement problem while the sensors’ positions are confined on the surface of the stator for the ease of installation. And then the gappy proper orthogonal decomposition (Gappy-POD) is used to reconstruct the magnetic field. An example of 2D surface-mounted PMSM with local demagnetization is presented. The results show that the Gappy-POD can reconstruct the magnetic field accurately with limited data. Comparing to the equal-spaced sensors, the sensor configuration determined by the proposed greedy algorithm provides better reconstruction performance. Furthermore, the sensor noise is considered and for the noise level below 0.5% the error is still low. At last, the k-nearest neighbor (KNN) method is proposed to classify the fault type and the support vector regression (SVR) method is proposed to predict the demagnetization remanence value. Both of them give a high prediction accuracy.
This paper presents dual finite element formulations for extraction of magnetic circuit (MC) parameter in electromagnetic devices. Analogous to electric circuit (EC), lumped parameters in MC, namely magnetic-resistance and magnetic-inductance, can be extracted by dual formulations employing scalar and vector potential variables. The dual formulations used in MC parameter extraction also shows the property of complementary energy bounds. Special treatments are developed to consider the case when the topology of both permeability domain and conductivity domain is nontrivial.