Discretization of a boundary-value problem with the eXtended Element-Free Galerkin (X-EFG) method yields an asymmetric EFG-type Saddle-Point (EFG-SP) problem that is difficult to solve numerically. As high-performance solvers for the problem, four types of the Asymmetric-version improved Variable-Reduction Methods (AiVRMs) are formulated. A numerical code is developed for solving asymmetric EFG-SP problems with four types of AiVRMs and, by means of the code, performances of the four methods are investigated numerically.

The multifunctional magnetically controlled transformer (MCT) organically integrates the magnetically controlled reactor with the distribution transformer. It has the advantages of fast and smooth adjustment of reactive power, and small overall floor area. Because of the topology of MCT is complex and affected by DC current, the analysis of electromagnetic loss characteristics on MCT iron cores is rather necessary. Firstly, this paper studied the structure and operating principle of MCT through the circuit analysis; Then, established the instantaneous core loss calculation model; Finally, quantitatively calculated the electromagnetic losses of MCT via finite element analysis (FEA), providing reference for its parameter design and heating verification.

The paper reports the progress to obtain accurate simulations of amplitude of EMAT signal from defects in metallic plates by means of FEM simulations. The simulations results are applied to a FEM model based on a two-dimensional coupled eddy-current (ECT) and ultrasound (UT) formulation, by using various numerical integration methods as Crank-Nicolson, Backward Euler and Newmark-beta method. The paper presents the methodology to acquire convergence of the large variation in the EMAT signal from a defect to the same final EMAT signal, not only in the defect amplitude but also in the shape of the time transient EMAT signal when using different methods. Accurate EMAT signals can be obtained by means of the combination of high order finite elements (FEM) and advanced numerical time integration, to reach the converged EMAT signal in the shortest time, opening the way to simulate accurate signals form faraway defects.

This paper presents a shape optimization of power inductor using actor-critic approach which is one of the reinforcement learning. The performance of the actor-critic (AC) approach is compared with that of the genetic algorithm. Comparing to GA, AC is more stably converged to an optimal solution with lower computational cost.