Search

The spread of high-performance personal computers, frequently equipped with powerful Graphic Processing Units (GPUs), has raised interest in a set of techniques that are able to extract models of electromagnetic phenomena (and devices) directly from available examples of desired behavior. Such approaches are collectively referred to as Machine Learning (ML). A typical representative ML approach is the so-called "Neural Network" (NN). Using such data-driven models allows the evaluation of the output in a much shorter time when a theoretical model is available, or allows the prediction of the behavior of the systems and devices when no theoretical model is available. With reference to a simple yet representative benchmark electromagnetic problem, some of the possibilities and pitfalls of the use of NNs for the interpretation of measurements (inverse problem) or to obtain required measurements (optimal design problem) are discussed. The investigated aspects include the choice of NN model, the generation of the dataset(s), and the selection of hyper-parameters (hidden layers, training paradigm). Finally, the capabilities in the handling of ill-posed problems are critically revised. [ABSTRACT FROM AUTHOR]

Real-time monitoring of hyperthermia treatments for liver cancers by noninvasive techniques was investigated in the last two decades. Among the proposed techniques, Electrical Impedance Tomography (EIT) arises particular interest for its low cost and sensitivity to tissues parameter changes to be detected. In this paper, an approach based on the intensive but critical exploitation of prior information is proposed and a preconditioning method of the impedance measurements based on a sensitivity analysis is assessed. Index Terms--Biomedical imaging, biomedical monitoring, impedance tomography, radio frequency hyperthermia.

Magneto-encephalography is a promising technique for the in vivo monitoring of brain activity. The information about neural activity is reconstructed from the electrical current map of the brain, which is in turn deduced, via an inverse problem resolution, from the magnetic field map measured by superconductive quantum interference devices placed around the head. In this paper, a possible strategy to improve the detection ability of the procedure by optimally designing the representation system is proposed and discussed. Index Terms--Electromagnetic modeling of biological systems, inverse problems, source identification.

In electromagnetic design, to counteract the effect of mechanical tolerances and/or assembly inaccuracies on the performance of 'optimized' devices, robust design techniques must be adopted. One of the most effective approaches takes advantage from statistical analysis to improve average performance rather than nominal one during the optimization process. Anyway, suitable measures must be taken to reduce the computational burden required by these strategies. In this paper a novel method for statistical design of electromagnetic power devices is presented, combining the strengths of existing methods while offering superior efficiency. It is inspired by Taguchi's approach and makes use of the unscented transformation. The computational requirements are considerably reduced compared to conventional statistical optimization methods. The proposed approach is implemented for the robust design of a magnetic resonance imaging superconducting device. Index Terms--Magnetic resonance imaging (MRI), optimization methods, robust design, unscented transformation.

The behavior of high temperature superconductor (HTS) coils is ruled by thermal and electromagnetic interacting phenomena. In this paper, a three-dimensional modeling of the quench propagation in HTS coils is discussed and numerically modeled. As a matter of fact, in spite of the axially symmetric design, a simple two-dimensional modelization may reveal too rough, particularly from the thermal point of view. In addition, the possibility of representing the complex structure of the coil composite material by means of a simplified homogeneous and anisotropic model is discussed and the advantages in terms of computational burden are evaluated. Index Terms--Coupled problems, electromagnetic analysis, high temperature superconductor (HTS).

A statistical approach is proposed for assessing the effects of manufacturing tolerances on the field homogeneity of Magnetic Resonance Imaging (MRI) magnets.

The research in the thermonuclear fusion area has entered the design phase of ITER, the first experimental Tokamak reactor. In the ITER device, the plasma shape and position are controlled by an active system on the basis of a real-time estimate of the plasma magnetic boundary. This identification process must be prompt and accurate enough to satisfy the tight control requirements. In this paper, a brief discussion about the reconstruction of the plasma boundary by means of the 'equivalent currents' (EC) method, in presence of conducting structures, will be presented. Some hints about a possible way to include the differential equations describing the vessel eddy currents will be illustrated. Index Terms--Eddy currents, nuclear fusion, plasma identification.

The critical current in multistrands superconducting (SC) cables is usually lower than expected due to current unbalance among strands and early current sharing. It is then crucial to be able to detect the current distribution in SC cables to assess their stability properties. Due to the ill-conditioning of the problem, the reconstruction accuracy may prove very sensitive to the measurement noise and to the uncertainty on the numerical data. In this work, a possible allocation strategy for the magnetic probes is discussed, aimed at the optimization of the reconstruction process. Index Terms--Identification, magnetic probes, superconducting (SC) cables.

Z-pinches are well-known plasma devices recently proposed as magnetic lenses. This focusing application requires an accurate modelling of the magnetic field distribution in the plasma discharge dynamics. It is a typical electromagnetic coupled problem, involving mechanical, thermal and circuital aspects, that have all to be properly modelled and coupled together. In this paper a model of the device, including energy balance and circuit coupling, is outlined, and its application to plasma discharges is discussed. Index terms - Magnetic fields, plasma pinch, numerical analysis, simulation, motion analysis, MHD, shock waves, particle beams.

A neural network model is proposed to treat inverse problems in electromagnetics, which includes wavelet functions to improve local approximation capabilities. This processor couples the advantages of an interpretation of the problem based on 'features' to the accuracy derived from using wavelets where local corrections are needed. The combined model allows to cope with singularities of the mapping and to slightly modify the mapping in real time. The detection and characterization of a circular defect in a conducting plate by using eddy current testing is shown to take advantage from the proposed approach in a test case, when unforeseen disturbances are present. Index terms - Neural Networks, Inverse Problems, Wavelet Decomposition, Eddy Current Testing

The knowledge and the control of the magnetic field profile is of major importance in the application of plasma discharges as magnetic lenses for charged particles beams, because of the complex dynamics of the plasma. Since conventional field measurements are quite difficult for high current plasma discharges, we propose an inverse formulation for the identification of the field profile directly from the focusing data. The problem is shown to be ill-posed, in general; therefore a regularisation is needed. This is obtained, first, by means of a proper restriction of the observation space; second, by using an over determined data set, leading to a pseudo-inversion procedure. The numerical solution is illustrated, showing the procedure to be robust, at least for the practical applications range. In particular it is shown how the proposed method is effective in determining the amount of distortion of the field profile as compared to the reference linear one. Index terms - Magnetic fields, particle beams, optimization, inverse problems.

The aim of this paper is to present a Lagrangian unified approach to shape-dependent electromagnetic problems (equilibrium, dynamic coupled, shape inverse, trajectory inverse problems). The main features of this approach are illustrated, and a comparison with the usual Eulerian point of view is made, from both the analytical and the numerical side, showing several advantages of the Lagrangian approach. A simple analytical example is presented in order to help comprehension and to show the effectiveness of the method. In addition, some practical applications are described. Index Terms - Coupled problems, electromagnetic analysis, finite element methods, inverse problems, mechanical systems, numerical analysis, optimization methods, velocity effects.