Raffaele Martone

A neural network approach for the solution of electric and magnetic inverse problems

Multilayer neural networks, trained via the back-propagation rule, are proved to provide an efficient means for solving electric and/or magnetic inverse problems. The underlying model of the system is learned by the network by means of a dataset defining the relationship between input and output parameters. The merits of the method are illustrated in the light of three example cases. The first two samples deal with inverse electrostatic problems which are relevant for nondestructive testing applications. In a first problem, a boss on an earthed plane is identified on the basis of the map of potential produced by a point charge. In the second problem, the geometric parameters of an ellipsoid carrying an electric charge are identified. In both cases, database of simulated measurements has been generated thanks to the available analytical solutions. As a sample magnetic inverse problem, the identification of a circular plasma in a tokamak device from external flux measurements is carried out. The results achieved show that the method here proposed is promising for technically meaningful applications. >

Multimodal Signals: Cognitive and Algorithmic Issues

Interactive and Unsupervised Multimodal Systems.- Multimodal Human Machine Interactions in Virtual and Augmented Reality.- Speech through the Ear, the Eye, the Mouth and the Hand.- Multimodality Issues in Conversation Analysis of Greek TV Interviews.- Representing Communicative Function and Behavior in Multimodal Communication.- Using the iCat as Avatar in Remote Meetings.- Using Context to Disambiguate Communicative Signals.- Modeling Aspects of Multimodal Lithuanian Human - Machine Interface.- Using a Signing Avatar as a Sign Language Research Tool.- Data Fusion at Different Levels.- Voice Technology Applied for Building a Prototype Smart Room.- Towards Facial Gestures Generation by Speech Signal Analysis Using HUGE Architecture.- Multi-modal Speech Processing Methods: An Overview and Future Research Directions Using a MATLAB Based Audio-Visual Toolbox.- From Extensity to Protensity in CAS: Adding Sounds to Icons.- Statistical Modeling of Interpersonal Distance with Range Imaging Data.- Verbal and Nonverbal Communication Signals.- How the Brain Processes Language in Different Modalities.- From Speech and Gestures to Dialogue Acts.- The Language of Interjections.- Gesture and Gaze in Persuasive Political Discourse.- Content in Embedded Sentences.- A Distributional Concept for Modeling Dialectal Variation in TTS.- Regionalized Text-to-Speech Systems: Persona Design and Application Scenarios.- Vocal Gestures in Slovak: Emotions and Prosody.- Spectrum Modification for Emotional Speech Synthesis.- Comparison of Grapheme and Phoneme Based Acoustic Modeling in LVCSR Task in Slovak.- Automatic Motherese Detection for Face-to-Face Interaction Analysis.- Recognition of Emotions in German Speech Using Gaussian Mixture Models.- Electroglottogram Analysis of Emotionally Styled Phonation.- Emoticonsciousness.- Urban Environmental Information Perception and Multimodal Communication: The Air Quality Example.- Underdetermined Blind Source Separation Using Linear Separation System.- Articulatory Synthesis of Speech and Singing: State of the Art and Suggestions for Future Research.- Qualitative and Quantitative Crying Analysis of New Born Babies Delivered Under High Risk Gestation.- Recognizing Facial Expressions Using Model-Based Image Interpretation.- Face Localization in 2D Frontal Face Images Using Luminosity Profiles Analysis.

Optimal magnetic probes location for current reconstruction in multistrands superconducting cables

The critical current in multistrand superconducting (SC) cables is usually lower than expected due to current imbalance among strands and early current sharing. It is thus 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 measurement noise and to the uncertainty of the numerical data. In this work, a possible allocation strategy for the magnetic probes is discussed, aimed at optimization of the reconstruction process.

Coupled Three Dimensional Numerical Calculation of Forces and Stresses on the End Windings of Large Turbo Generators via Integral Formulation

A novel numerical approach to calculate the time evolution of the three dimensional distribution of the magnetic field and forces in the end winding regions of large turbine generators is presented. The proposed approach is based on an integral formulation for nonlinear magnetostatic problems. Its main advantage is the reduction of the discretization to only the conductors and magnetic materials. In this paper the solution of a coupled magnetostructural problem consisting in the calculation of the mechanical stresses and deformations caused by the electrodynamic forces is presented. The analysis is based on a time stepping simulation where the currents are derived from the integration of a lumped parameter model.

A genetic algorithm approach to the design of split coil magnets for MRI

Innovative, open magnet designs for Magnetic Resonance Imaging (MRI) of parts of human body have been recently proposed. These open configurations require notable skill in the design phase to obtain satisfactory performance in terms of compactness and field homogeneity; optimization techniques thus seems adequate to assist the designer of such devices. A novel optimization scheme, based on a genetic algorithm (GA) has been implemented for the resolution of this problem. The GA seems well suited for this applications thanks to its ability in exploring the whole design parameters space, not being trapped by local minima, but some improvements have been implemented, namely floating point parameter representation, introduction of a directional mutation operator and adaptive probability of operators action.

A Lagrangian approach to shape inverse electromagnetic problems

This paper presents a completely Lagrangian approach which is able to solve shape inverse (optimization and identification) electromagnetic problems. A first order deterministic minimization technique combined with Lagrangian finite element approach is used. Theoretical and numerical advantages and drawbacks of such a method are discussed, and some applications are presented.

Stochastic handling of tolerances in robust magnets design

Due to construction tolerances, the performances of actual electromagnetic devices differ from those of the nominal design. A strategy for the search of design solutions robust against construction tolerances is presented. The approach is based on the statistical analysis of the tolerances impact on the design objective function to be evaluated. The application of the proposed strategies to the optimal design of magnetic resonance imaging magnets by a genetic algorithm shows the possibility to direct the search toward robust solutions.

An error based approach to the solution of full Maxwell equations

Aim of this paper is to extend the error based approach to the study of general electromagnetic problems in 3D geometries in which the displacement currents may not be neglected. The unknown variables are the three-component vector potential A and W defined as the time integrals of -E and H, respectively. These potentials are constrained to satisfy initial, boundary and interface conditions. Since in this way the Maxwell equations are automatically satisfied, the solution is obtained via minimization of a global error functional which approaches zero when the constitutive equations are satisfied. >

Effectiveness in 3-D Magnetic Field Evaluation of Complex Magnets

Numerical high-accuracy evaluation of 3-D magnetic field generated by complex magnetic structures can be a quite demanding task. In order to save the computational effort, while preserving the accuracy level, a number of simple magnetic elements able to model complex 3-D structures are available. In this paper, a critical comparison of different modeling has been carried out in the perspective of a graphical processing unit implementation.

Tolerance Analysis of NMR Magnets

Due to manufacturing and assembly variations, there are tolerances associated with the nominal dimensions of complex electromagnetic devices. A possible approach to take into account such tolerances in the early design phases is the “worst case tolerance analysis,” whose goal is to determine the effect of the largest variations on assembly dimensions on the product performance. On the other hand, when many tolerances must be considered, locating the worst case with combinatorial effects could get very time consuming, since performance assessment of such devices may require numerical field analysis. In this paper, a fast approach based on the performance sensitivity with respect to design parameters is presented, and applied to the tolerance analysis of nuclear magnetic resonance (NMR) magnets.